MPEG-2 Technical (and sometimes political) Frequently Asked Questions
(FAQ) list.
Copyright 1994 by Chad Fogg (cfogg@netcom.com)
Draft 3.3 (May 10, 1994)
1. MPEG is a DCT based scheme, right?
2. What does the MPEG video syntax feature that codes video efficiently?
3. What does the syntax provide for error robustness?
4. What is the significance of each layer in MPEG video ?
5. How does the syntax facilitate parallelism?
6. I hear the encoder is not part of the standard?
7. Are some encoders better than others?
8. Can MPEG-1 encode higher sample rates than 352 x 240 x 30 Hz ?
9. What are Constrained Parameters Bitstreams (CPB) for video?
10. Why is Constrained Parameters so important?
11. Who uses constrained parameters bitstreams?
12. Are there ways of circumventing constrained parameters bitstreams for SIF
class applications and decoders ?
13. Are there any other conformance points like CPB for MPEG-1?
14. What frame rates are permitted in MPEG?
15. Special prediction switches for MPEG-2
16. What is MPEG-2 Video Main Profile and Main Level?
17. Does anybody actually use the scalability modes?
18. What's the difference between Field and Frame pictures?
19. What do B-pictures buy you?
20. Why do some people hate B-frames?
21. Why was the 16x16 area chosen?
22. Why was the 8x8 DCT size chosen?
23. What is motion compensated prediction, and why is it a pain?
24. What are the various prediction modes in MPEG-2?
24.1 Frame:
24.2 Field predictions in frame-coded pictures:
24.3 Field predictions in field-coded pictures:
24.4 16x8 predictions in field-coded pictures:
24.5 Dual Prime prediction in frame and field-coded pictures
24.6 Field and frame organized macroblocks:
25. How do you tell a MPEG-1 bitstream from a MPEG-2 bitstream?
26. What is the reasoning behind MPEG syntax symbols?
27. Why bother to research compressed video when there is a standard?
28. Where can I get a copy of the latest MPEG-2 draft?
29. What are the latest working drafts of MPEG-2 ?
30. What is the latest version of the MPEG-1 documents?
31. What is the evolution of ISO standard documents?
32. Where is a good introductory paper to MPEG?
33. What are some journals on related MPEG topics ?
34. Is there a book on MPEG video?
35. Is it MPEG-2 (Arabic numbers) or MPEG-II (roman)?
36. What happened to MPEG-3?
37. What is MPEG-4?
38. What are the scaleable modes of MPEG-2?
39. Why MPEG-2? Wasn't MPEG-1 enough?
40. What did MPEG-2 add to MPEG-1 in terms of syntax/algorithms ?
41. How do MPEG and JPEG differ?
42. How do MPEG and H.261 differ?
43. Is H.261 the de facto teleconferencing standard?
44. What is the TM rate control and adaptive quantization technique ?
45. How does the TM work?
46. What is a good motion estimation method, then?
47. Is exhaustive search "optimal" ?
48. What are some advanced encoding methods?
49. Is so-and-so really MPEG compliant ?
50. What are the tell-tale MPEG artifacts?
51. Where are the weak points of MPEG video ?
52. What are some myths about MPEG?
53. What is the color space of MPEG?
54. Don't you mean 4:1:1 ?
55. Why did MPEG choose 4:2:0 ? Isn't 4:2:2 the standard for TV?
56. What is the precision of MPEG samples?
57. What is all the fuss with cositing of chroma components?
58. How would you explain MPEG to the data compression expert?
59. How does MPEG video really compare to TV, VHS, laserdisc ?
60. What are the typical MPEG-2 bitrates and picture quality?
61. At what bitrates is MPEG-2 video optimal?
62. Why does film perform so well with MPEG ?
63. What is the best compression ratio for MPEG ?
64. Can MPEG be used to code still frames?
65. Is there an MPEG file format?
66. What are some pre-processing enhancements ?
67. Why use these "advanced" pre-filtering techniques?
68. What about post-processing enhancements?
69. Can motion vectors be used to measure object velocity?
70. How do you code interlaced video with MPEG-1 syntax?
71. Is MPEG patented?
72. How many cable box alliances are there?
73. Will there be an MPEG video tape format?
74. Where will be see MPEG in everyday life?
75. What is the best compression ratio for MPEG ?
76. Is there a MPEG CD-ROM format?
1. MPEG is a DCT based scheme, right?
The DCT and Huffman algorithms receive the most press coverage (e.g. MPEG is
a DCT based scheme with Huffman coding), but are in fact fairly
insignificant. The variety of coding modes signaled to the decoder as
context-dependent side information are chiefly responsible for the efficiency
of the MPEG syntax.
2. What does the MPEG video syntax feature that codes video efficiently?
A. Here are some of the statistical conditions and their syntax counterparts.
Occlusion: forward, backwards, or bi-directional temporal prediction in B
pictures.
Smooth optical flow fields: variable length coding of 1-D prediction errors
for motion vectors.
Spatial correlation beyond 8x8 sample block boundaries: 1-D prediction of
DC coefficients in consecutive group intra-coded macroblocks.
High temporal correlation: variable on/off coding of prediction error at
the macroblock (no-coding) or individual block (coded block pattern) level.
Temporal de-correlation: forward, backwards, or bidirectional prediction.
Content dependent quality: locally adaptive quantization
Temporal prediction accuracy: "half-pel" sample accuracy.
High locally correlated signal refresh pictures (I picture) and prediction
errors: DCT
Subjective coding: location-dependent quantization of DCT coefficients.
3. What does the syntax provide for error robustness?
1. Byte-aligned start codes in the coded bitstream.
2. End of block codes in coded blocks.
3. Slices.
4. slice_vertical_position embedded as sub-field within slice start codes.
5. slices commencing at regular locations in picture (MPEG-2)
4. What is the significance of each layer in MPEG video ?
Sequence:
Set of pictures sharing same sampling dimensions, bit rate,
chromaticy (MPEG-1), quantization matrices (MPEG-1 only).
Group of Pictures:
Random access point giving SMPTE time code within sequence.
Guaranteed to start with an I picture.
Picture:
Samples of a common plane -- "captured" from the same time instant.
Slice:
Error resynchronization unit of macroblocks.
At the commencement of a slice, all inter-macroblock coding
dependencies are reset. Likewise, all macroblocks within a common slice
can be dependently coded.
Macroblock:
Least common multiple of Y, Cb, Cr 8x8 blocks in 4:2:0 sampling
structure.
For MPEG-1, the smallest granularity of temporal prediction.
Block:
Smallest granularity of spatial decorrelation.
5. How does the syntax facilitate parallelism?
A. For MPEG-1, slices may consist of an arbitrary number of macroblocks.
The coded bitstream must first be mapped into fixed-length elements before
true parallelism in a decoder application can be exploited. Further, since
macroblocks have coding dependencies on previous macroblocks within the same
slice, the data hierarchy must be pre-processed down to the layer of DC DCT
coefficients. After this, blocks may be independently inverse transformed
and quantized, temporally predicted, and reconstructed to buffer memory.
Parallelism is usually more of a concern for encoders. Macroblock motion
estimation and some rate control stages can be processed independently. An
encoder also has the freedom to choose the slice structure.
6. I hear the encoder is not part of the standard?
A. The encoder rests just outside the normative scope of the standard,
as long as the bitstreams it produces are compliant. The decoder,
however, is almost deterministic: a given bitstream should
reconstruct to a unique set of pictures. Statistically speaking, an
occasional error of a Least Significant Bit is permitted as a
result of the fact that the IDCT function is the only non-normative
stage in the decoder (the designer is free to choose among many DCT
algorithms and implementations). The IEEE 1180 test referenced in
Annex A of the MPEG-1 and MPEG-2 specifications spells out the
statistical mismatch tolerance between the Reference IDCT, which
uses 64-bit floating point accuracy, and the Test IDCT.
7. Are some encoders better than others?
A. Yes. For example, the range over which a compensated prediction
macroblock is searched for has a great influence over final picture
quality. At a certain point a very large range can actually become
detrimental (it may encourage large differential motion vectors).
Practical ranges are usually between +/- 15 and +/- 32. As the
range doubles, for instance, the search area quadruples.
8. Can MPEG-1 encode higher sample rates than 352 x 240 x 30 Hz ?
A. Yes. The MPEG-1 syntax permits sampling dimensions as high as 4095 x
4095 x 60 frames per second. The MPEG most people think of as "MPEG-
1" is actually a kind of subset known as Constrained Parameters
bitstream (CPB).
9. What are Constrained Parameters Bitstreams (CPB) for video?
A. MPEG-1 CPB are a limited set of sampling and bitrate parameters
designed to normalize decoder computational complexity, buffer size, and
memory bandwidth while still addressing the widest possible range of
applications. The parameter limits were intentionally designed so that a
decoder implementation would need only 4 Megabits of DRAM.
Parameter Limit
-------------- ---------------------------
pixels/line 704
lines/picture 480 or 576
pixels*lines 352*240 or 352*288
picture rate 30 Hz
bit rate 1.862million bits/sec
buffer size 40 Kilobytes (327,680 bits)
The sampling limits of CPB are bounded at the ever popular SIF rate:
396 macroblocks (101,376 pixels) per picture if the picture rate is
less than or equal to 25 Hz, and 330 macroblocks (84,480 pixels) per
picture if the picture rate is 30 Hz. The MPEG nomenclature loosely
defines a "pixel" or "pel" as a unit vector containing a complete
luminance sample and one fractional (0.25 in 4:2:0 format) sample from
each of the two chrominance (Cb and Cr) channels. Thus, the
corresponding bandwidth figure can be computed as:
352 samples/line x 240 lines/picture x 30 pictures/sec x 1.5 samples/pixel
or 3.8 Ms/s (million samples/sec) including chroma, but not including
blanking intervals. Since most decoders are capable of sustaining
VLC decoding at a faster rate than 1.8 Mbit/sec, the coded video bitrate
has become the most often waived parameter of CPB. An encoder which
intelligently employs the syntax tools should achieve SIF quality saturation
at about 2 Mbit/sec, whereas an encoder producing streams containing
only I (Intra) pictures might require as much as 4 Mbit/sec to achieve the
same video quality.
10. Why is Constrained Parameters so important?
A. It is an optimum point that allows (just barely) cost effective VLSI
implementations in 1992 technology (0.8 microns). It also implies a
nominal guarantee of interoperability for decoders and encoders. Since
CPB is a canonical conformance point, MPEG devices which are not capable
of meeting SIF rates are usually not considered to be true MPEG.
11. Who uses constrained parameters bitstreams?
A. Applications which are focused on CPB are Compact Disc (White Book or CD-
I) and computer video applications. Set-top TV decoders fall into a higher
sampling rate category known as CCIR 601 or Broadcast rate.
12. Are there ways of circumventing constrained parameters bitstreams for SIF
class applications and decoders ?
A. Yes, some. Remember that CPB limits pictures by macroblock count.
416 x 240 x 24 Hz sampling rates are still within the constraints, but this
would only be of benefit in NTSC (240 lines/field) displays. Deviating from
352 samples/line could throw off many decoder implementations which possess
limited horizontal sample rate conversion abilities. Some decoders do in fact
include a few rate conversion modes, with a filter usually implemented via
binary taps (shifts and adds). Likewise, the target sample rates are usually
limited or ratios (e.g. 640, 540, 480 pixels/line, etc.). Future MPEG
decoders will likely include on-chip arbitrary sample rate converters,
perhaps capable of operating in the vertical direction (although there is
little need of this in applications using standard TV monitors, with the
possible exception of windowing in cable box graphical user interfaces).
13. Are there any other conformance points like CPB for MPEG-1?
A. Undocumented ones, yes. A second generation of decoder chips emerged on
the market about 1 year after the first wave of SIF-class decoders. Both
LSI Logic and SGS-Thomson introduced CCIR 601 class MPEG-1 decodersto fill in
the gap between canonical MPEG-1 and the emergence of MPEG-2. Under non-
disclosure agreement, C-Cube had the CL-950.
14. What frame rates are permitted in MPEG?
A. A limited set is available for the choosing in MPEG-1, although "tricks"
could be played with Systems-layer Time Stamps to convey non-standard rates.
The set is: 23.976 Hz (3-2 pulldown NTSC), 24 Hz (Film), 25 Hz (PAL/SECAM or
625/60 video), 29.97 (NTSC), 30 Hz (drop-frame NTSC or component 525/60), 50
Hz (double-rate PAL), 59.97 Hz (double rate NTSC), and 60 Hz (double-rate
drop-frame NTSC/component 525/60 video).
15. Special prediction switches for MPEG-2
MPEG-2 sequence
/ \
progressive interlaced sequence
sequence / \
Field picture Frame picture
/ \
Frame or field pred. Frame MB prediction only
/ \
Field dct Frame dct
16. What is MPEG-2 Video Main Profile and Main Level?
A. MPEG-2 Video Main Profile and Main Level is analogous to MPEG-1's CPB,with
sampling limits at CCIR 601 parameters (720 x 480 x 30 Hz). Profiles limit
syntax (i.e. algorithms), whereas Levels place limits on coding parameters
(sample rates, frame dimensions, coded bitrates, etc.). Together, Video Main
Profile and Main Level (abbreviated as MP@ML) normalize complexity within
feasible limits of 1994 VLSI technology (0.5 micron), yet still meet the
needs of the majority of application users.MP@ML is the conformance point for
most cable and satellite systems.
Profiles
======
Simple: I and P pictures only. 4:2:0 sampling ratio. 8,9, or 10 bits DC
precision.
Main: I, P, and B pictures. Dual Prime with no B-pictures only. 4:2:0
sampling ratio. 8, 9, or 10 bits sample precision.
SNR profile:
Spatial profile:
High: 8,9,10, or 11 bits sample precision. 4:2:2 and 4:4:4 sampling ratio.
Level
====
Simple: SIF video rate (3.041280 Mhz), 4 Mbit/sec, 0.489472 Mbit VBV
buffer, 64 vertical in frame, 32Vertical in field, 1:7 fcode hor.
Main: CCIR 601 video rate (10.368 Mhz), 15 Mbit/sec, 1.835008 Mbit VBV
buffer, 128 V in frame, 64 V in field, 1:8 f_code Hor.
High 1440: 1440 x 1152 x 30 Hz (47.0016 Mhz), 60 Mbit/sec. 7.340032 Mbit
VBV buffer, 128 V in Fe, 1:9 fcode H.
High: 1920 x 1152 x 30 Hz (62.6688 Mhz), 80 Mbit/sec. 9.787392 Mbit VBV
buffer.
1:9 fcode H
17. Does anybody actually use the scalability modes?
A. At this time, scalability has found itself a limited number of
applications, although research is definitely underway for its use in HDTV.
Experiments have been demonstrated in Europe where, for example, PAL-rate
video (720 x 576 x 25 fps) is embedded in the same stream as HDTV rate video
(1440 x 1152 x 25 fps). The Nov. 1992 VADIS experiment divided the base layer
(PAL) and enhancement into 4 and 16 Mbit/sec channels, respectively. The U.S.
Grand Alliance favors HDTV simulcasting (separate NTSC analog and digital
HDTV broadcasts). Temporal scalability is the pet scalability mode as the
possible future solution for coding 60 Hz progressive sequences while
maintaining backwards compatibility with early-wave equipment (e.g. 1920 x
1080 x 30 Hz displays) . To elaborate, the first wave receivers of the late
1990's would be limited to 62at 0 Hz interlaced/30 Hz progressive HDTV
decoders. Essentially, 60 interlaced fields would be coded in a, for
example, 16 Mbit/sec stream in 1996, and when VLSI processes shift another
thousand or so angstroms down the wavelength scale, an 8 Mbit/sec enhancement
layer containing the coded "high pass" between 60 Hz progressive and 60 Hz
interlaced would be simulcasted or multiplexed. Several corporate mouths
have been known to water at the mention of charging the quality conscious
subscriber an extra fee for the enhancement layer.
18. What's the difference between Field and Frame pictures?
A. A frame-coded picture consists of samples from both even and odd fields.
A
frame picture is coded in progressive order (an even line, then an odd line,
etc.) and in the case of MPEG-2, may optionally switch between field and
frame order on a macroblock basis. The Display Process, which is *almost*
completely outside the scope of the MPEG specification, can chose to re-
interlace the picture by displaying the odd and even lines at different times
(16 milliseconds apart for 60 Hz displays). In fact, most pictures,
regardless of whether they were coded as a Field or Frame, end up being
displayed interlaced due to the fact that most TV sets are interlaced.
19. What do B-pictures buy you?
A. Since bi-directional macroblock predictions are an average of two
macroblock areas, noise is reduced at low bit rates (like a 3-D filter, if
you will). At nominal MPEG-1 video (352 x 240 x 30, 1.15 Mbit/sec) rates, it
is said that B-frames improves SNR by as much as 2 dB. (0.5 dB gain is
usually considered worth-while in MPEG). However, at higher bit rates, B-
frames become less useful since they inherently do not contribute to the
progressive refinement of an image sequence (i.e. not used as prediction by
subsequent coded frames). Regardless, B-frames are still politically
controversial.
B pictures are interpolative in two ways: 1. predictions in the bi-
directional macroblocks are an average from block areas of two pictures 2. B
pictures fill in or interpolate the 3-D video signal over a 33 or 25
millisecond picture period without contributing to the overall signal quality
beyond that immediate point in time. In other words, a B pictures,
regardless of its internal make-up of macroblock types, has a life limited to
its immediate self. As mentioned before, its energy does not propagate into
other frames. In a sense, bits spent on B pictures are wasted.
20. Why do some people hate B-frames?
A. Computational complexity, bandwidth, delay, and picture buffer size are
the four B-frame Pet Peeves. Computational complexity in the decoder is
increased since a some macroblock modes require averaging between two
macroblocks.
Worst case, memory bandwidth is increased an extra 15.2 MByte/s (4:2:0 601
rates, not including any half pel or page-mode overhead) for this extra
prediction. An extra picture buffer is needed to store the future prediction
reference (bi-directionality). Finally, extra delay is introduced in
encoding since the frame used for backwards prediction needs to be
transmitted to the decoder before the intermediate B-pictures can be decoded
and displayed.
Cable television (e.g. -- more like i.e.-- General Instruments) have been
particularly adverse to B-frames since, for CCIR 601 rate video, the extra
picture buffer pushes the decoder DRAM memory requirements past the magic 8-
Mbit (1 Mbyte) threshold into the evil realm of 16 Mbits (2 Mbyte)....
although 8-Mbits is fine for 352 x 480 B picture sequence. However, cable
often forgets that DRAM does not come in convenient high-volume (low cost) 8-
Mbit packages as does the friendly 4-Mbit and 16-Mbit. In a few years, the
cost difference between 16 Mbit and 8 Mbit will become insignificant compared
to the bandwidth savings gain through higher compression. For the time
being, some cable boxes will start with 8-Mbit and allow future drop-in
upgrades to the full 16-Mbit.
21. Why was the 16x16 area chosen?
A. The 16x16 area corresponds to the Least Common Multiple (LCM) of 8x8
blocks, given the normative 4:2:0 chroma ratio. Starting with medium
size images, the 16x16 area provides a good balance between side
information overhead & complexity and motion compensated prediction
accuracy. In gist, 16x16 seemed like a good trade-off.
22. Why was the 8x8 DCT size chosen?
A. Experiments showed little improvements with larger sizes vs. the
increased complexity. A fast DCT algorithm will require roughly double
the arithmetic operations per sample when the transform point size is
doubled. Naturally, the best compaction efficiency has been demonstrated
using
locally adaptive block sizes (e.g. 16x16, 16x8, 8x8, 8x4, and 4x4) [See
Baker and Sullivan]. Naturally, this introduces additional side information
overhead and forces the decoder to implement programmable or hardwired
recursive DCT algorithms. If the DCT size becomes too large, then more edges
(local discontinuities) and the like become absorbed into the transform
block, resulting in wider propagation of Gibbs (ringing) and other phenomena.
Finally, with larger transform sizes, the DC term is even more critically
sensitive to quantization noise.
23. What is motion compensated prediction, and why is it a pain?
A. MCP in the decoder can be thought of as having four stages:
1. Motion vector computation
2. Prediction retrieval
various predictions are 16x16, 16x8, 8x4, 8x8 plus any half-pel
overhead (e.g. 17x16, 17x17, etc).
3. Filtering
3.1 Forming half-pel predictions through bi-linear interpolation.
3.2 Averaging two predictions together (B macroblocks, Dual Prime)
4. Combination and ordering
4.1 combining 1 or 2 predictions from stage three into upper and
lower halves (16 x 8, field in frame)
4.2 interleaving or grouping together odd and even lines in frame
picture predictions.
The final, combined prediction is always a 16x16 block of luminance and
8x8 block of chrominance, just like we experience in MPEG-1.
A single motion vector can be associated with each source, hence a macroblock
can have as many as 4 motion vectors.
24. What are the various prediction modes in MPEG-2?
24.1 Frame:
Predictions are formed from a 16 x 16 pixel area in a previously
reconstructed frame. Identical to MPEG-1. There can be only one source in
forward or backward predicted macroblocks, and two sources in bi-directional
macroblocks. The prediction frame itself may have been coded as either a
frame or two fields, however once a frame is reconstructed, it is simply a
frame as far as future predictions are concerned.
24.2 Field predictions in frame-coded pictures:
Separate predictions are formed for the top (8 lines from field 1)and bottom
(8 lines from field 2) portions of the macroblock. A total of two motion
vectors in forward or backward predictions, four in bi-directional.
24.3 Field predictions in field-coded pictures:
Predictions are formed from the two most recently decoded fields. Prediction
sizes are 16x16, however the 16 lines have a corresponding projection onto a
16x32 pixel area of a frame. One motion vector for forward or backward
predictions, and two for bi-directional.
24.4 16x8 predictions in field-coded pictures:
Like field macroblocks in frame-coded pictures, the upper and lower 8 lines
in this macroblock mode can have different predictions (hence two motion
vectors). This mode compensates for the reduced temporal prediction
precision of field picture macroblocks (a result of the fact that fields
inherently possess half the number of lines that frames do). The field
prediction area projected onto a frame is restored to 16 lines. 2 motion
vectors for backwards or forwards, 4 for bi-directional.
24.5 Dual Prime prediction in frame and field-coded pictures
Predictions for the current macroblock are formed from the average of two 16
x 8 line areas from the two most recently decoded fields. Dual Prime was
devised as an alternative for B pictures in low delay applications, but still
offers many of the signal
quality benefits of B-pictures. Dual Prime requires one less prediction
picture buffer, but still retains the same instantaneous prediction bandwidth
of a B picture system. As an alternative to coding separate motion vectors
for each of the upper and lower 16x8 areas, a full motion vector is sent for
the first area, and a +1, 0, or -1 differential vector (variable length
coded) is specified for the second prediction area. A macroblock will have
total of two full motion vectors and two differential vectors in frame-coded
pictures. Due to the prediction bandwidth overhead, Main Profile restricts
the use of Dual Prime prediction to P picture sequences only. High Profile
permits use of Dual Prime in B pictures.
24.6 Field and frame organized macroblocks:
Originally intended as a cheaper means of achieving field-decorrelation in
frame-coded pictures without the fussy overhead of separate field prediction
estimates, the dct coefficients (quantized prediction error for a given
macroblock) may be organized into either a field or frame pattern.
Essentially this means that the prediction error for the combined 16x16
macroblock may be grouped into field or frame blocks. A bit in the macroblock
header (dct_type) indicates whether the upper and lower portions of the
macroblock are to be interleaved (frame organized) or remain separated (field
organized).
25. How do you tell a MPEG-1 bitstream from a MPEG-2 bitstream?
A. All MPEG-2 bitstreams must contain specific extension headers that
*immediately* follow MPEG-1 headers. At the highest layer, for example,
the MPEG-1 style sequence_header() is followed by sequence_extension()
exclusive to MPEG-2. Some extension headers are specific to MPEG-2 profiles.
For example, sequence_scalable_extension() is not allowed in Main Profile
bitstreams.
A simple program need only scan the coded bitstream for byte-aligned start
codes to determine whether the stream is MPEG-1 or MPEG-2.
26. What is the reasoning behind MPEG syntax symbols?
A. Here are some of the Whys and Wherefores of MPEG symbols:
Start codes
These 32-bit byte-aligned codes provide a mechanism for cheaply
searching coded bitstreams for commencement of various layers of video
without having to actually parse variable-length codes or perform any
decoder arithmetic. Start codes also provide a mechanism for
resynchronization in the presence of bit errors.
Coded block pattern (CBP --not to be confused with Constrained
Parameters!) When the frame prediction is particularly good, the
displaced frame difference (DFD, or prediction error) tends to be small,
often with entire block energy being reduced to zero after quantization.
This usually happens only at low bit rates. Coded block patterns
prevent the need for transmitting EOB symbols in those zero coded
blocks.
DCT_coefficient_first
Each intra coded block has a DC coefficient. With coded block patterns
signaling all possible combinations of all-zero valued blocks, the
dct_coef_first mechanism assigns a different meaning to the VLC codeword
that would otherwise represent EOB as the first coefficient.
End of Block:
Saves unnecessary run-length codes. At optimal bitrates, there tends to
be few AC coefficients concentrated in the early stages of the zig-zag
vector. In MPEG-1, the 2-bit length of EOB implies that there is an
average of only 3 or 4 non-zero AC coefficients per block. In MPEG-2
Intra (I) pictures, with a 4-bit EOB code, this number is between 9 and
16 coefficients. Since EOB is required for all coded blocks, its absence
can signal that a syntax error has occurred in the bitstream.
Macroblock stuffing
A genuine pain for VLSI implementations, macroblock stuffing was
introduced to maintain smoother, constant bitrate control in MPEG-1.
However, with normalized complexity measures and buffer management
performed a priori (pre-frame, pre-slice, and pre-macroblock) in the
MPEG-2 encoder test model, the need for such localized smoothing
evaporated. Stuffing can be achieved through virtually unlimited slice
start code padding if required. A good rule of thumb: if you find
yourself often using stuffing more than once per slice, you probably
don't have a very good rate control algorithm. Anyway, macroblock
stuffing is now illegal in MPEG-2, so don t start using it if you
already haven t.
MPEG's modified Huffman VLC tables
The VLC tables in MPEG are not Huffman tables in the true sense of
Huffman coding, but are more like the tables used in Group 3 fax. They
are entropy constrained, that is, non-downloadable and optimized for a
limited range of bit rates (sweet spots). With the exception of a few
codewords, the larger tables were carried over from the H.261 standard
of 1990. MPEG-2 added an "Intra table". Note that the dct_coefficient
tables assume positive/negative coefficient pmf symmetry.
27. Why bother to research compressed video when there is a standard?
A. Despite the worldwide standard, many areas remain open for research:
advanced encoding and pre-processing, motion estimation, macroblock
decision models, rate control and buffer management in editing
environments, etc. There's practically no end to it.
28. Where can I get a copy of the latest MPEG-2 draft?
A. Contact your national standards body (e.g. ANSI Sales in NYC for the
U.S., British Standards Institute in the UK, etc.). A number of private
organizations offer ISO documents.
29. What are the latest working drafts of MPEG-2 ?
A. MPEG-2 has reached voting document of the Draft International Standard for
:
Information Technology -- Generic Coding of Moving Pictures and
Associated Audio. Recommendation H.262, ISO/IEC Draft International Standard
13818-2. [produced March 25, 1994, not yet approved by voting process].
Audio is Part 1, Video Part 2, and Systems is Part 3. A committee draft for
Conformance (Part 4) is expected in Novemeber 1994, as well as the Technical
Report on Software Simulation (Part 5).
30. What is the latest version of the MPEG-1 documents?
A. Systems (ISO/IEC IS 11172-1), Video (ISO/IEC IS 11172-2), and Audio
(ISO/IEC IS 11172-3) have reached the final document stage. Part 4,
Conformance Testing, is currently DIS
31. What is the evolution of ISO standard documents?
A. In chronological order:
ISO/Committee notation Author's notation
--------------------------------------- -------------------------
Problem (unofficial first stage) Barroom Witticism
New work Item (NI) Napkin Item
New Proposal (NP) Need Permission
Working Draft (WD) We're Drunk
Committee Draft (CD) Calendar Deadlock
Draft International Standard (DIS) Doesn't Include Substance
International Standard (IS) Induced patent Statements
32. Where is a good introductory paper to MPEG?
A. Didier Le Gall, "MPEG: A Video Compression Standard for Multimedia
Applications," Communications of the ACM, April 1991, Vol.34, No.4, pp. 47-58
33. What are some journals on related MPEG topics ?
A.
IEEE Transactions on Consumer Electronics
IEEE Transactions on Broadcasting
IEEE Transactions on Circuits and Systems for Video Technology
Advanced Electronic Imaging
Electronic Engineering Times (EE Times -- more tabloid coverage. Unfortunate
columns by Richard Doherty)
IEEE Int'l Conference on Acoustics, Speech, and Signal Processing
(ICASSP)
International Broadcasting Convention (IBC)
Society of Motion Pictures and Television Engineers (SMPTE)
SPIE conference on Visual Communications and Image Processing
SPIE conference on Video Compression for Personal Computers
IEEE Multimedia [first edition Spring 1994]
34. Is there a book on MPEG video?
A. Yes, there will be a book published sometime in 1994 by the same authors
who brought you the JPEG book (Bill Pennebaker, Joan Mitchell). Didier Le
Gall will be an additional co-author, and will insure digressions into, e.g.
arithmetic coding aspects, be kept to a minimum :-)
35. Is it MPEG-2 (Arabic numbers) or MPEG-II (roman)?
A. Committee insiders most often use the Arabic notation with the hyphen,
e.g. MPEG-2. Only the most retentive use the official designation: Phase 2.
In fact, M.P.E.G. itself is a nickname. The official title is: ISO/IEC JTC1
SC29 WG11. The militaristic lingo has so far managed to keep the enemy
(DVI) confused and out of the picture.
ISO: International Organization for Standardization
IEC: International Electrotechnical Commission
JTC1: Joint Technical Committee 1
SC29: Sub-committee 29
WG11: Work Group 11 (moving pictures with... uh, audio)
36. What happened to MPEG-3?
A. MPEG-3 was to have targeted HDTV applications with sampling dimensions up
to 1920 x 1080 x 30 Hz and coded bitrates between 20 and 40 Mbit/sec. It was
later discovered that with some (compatible) fine tuning, MPEG-2 and MPEG-1
syntax worked very well for HDTV rate video. The key is to maintain an
optimal balance between sample rate and coded bit rate.
Also, the standardization window for HDTV was rapidly closing. Europe and
the United States were on the brink of committing to analog-digital
subnyquist hybrid algorithms (D-MAC, MUSE, et al). European all-digital
projects such as HD-DIVINE and VADIS demonstrated better picture quality with
respect to bandwidth using the MPEG syntax. In the United States, the
Sarnoff/NBC/Philips/Thomson HDTV consortium had used MPEG-1 syntax from the
beginning of its all-digital proposal, and with the exception of motion
artifacts (due to limited search range in the encoder), was deemed to have
the best picture quality of all three digital proponents. HDTV is now part of
the MPEG-2 High-1440 Level and High Level toolkit.
37. What is MPEG-4?
A. MPEG-4 targets the Very Low Bitrate applications defined loosely as
having sampling dimensions up to 176 x 144 x 10 Hz and coded bit rates
between 4800 and 64,000 bits/sec. This new standard would be used, for
example, in low bit rate videophones over analog telephone lines.
This effort is in the very early stages. Morphology, fractals, model
based, and anal retentive block transform coding are all in the
offering. MPEG-4 is now in the application identification phase.
Scaleable modes of MPEG-2
38. What are the scaleable modes of MPEG-2?
A. Scaleable video is permitted only in the High Profiles.
Currently, there are four scaleable modes in the MPEG-2 toolkit. These modes
break MPEG-2 video into different layers (base, middle, and high layers)
mostly for purposes of prioritizing video data. For example, the high
priority channel (bitstream) can be coded with a combination of extra error
correction information and/or increased signal strength (i.e. higher Carrier-
to-Noise ratio or lower Bit Error Rate) than the lower priority channel. For
example, in HDTV, the high priority bitstream (720 x 480) can be decoded
under noise conditions were the lower priority (1440 x 960) cannot. This is
part of the "graceful degradation concept. Breaking a video signal into two
streams (base and enhancements) has a penalty, however. Usually less than
1.5 dB.
Another purpose of salability is complexity division. A standard TV set need
only decode the 720 x 480 channel, thus requiring a less expensive decoder
processor than a TV set wishing to display 1440 x 960. This is known as
simulcasting.
A brief summary of the MPEG-2 video scalability modes:
Spatial Scalablity-- Useful in simulcasting, and for feasible software
decoding of the lower resolution, base layer. This spatial domain
method codes a base layer at lower sampling dimensions (i.e.
"resolution") than the upper layers. The upsampled reconstructed lower
(base) layers are then used as prediction for the higher layers.
Data Partitioning-- Similar to JPEG's frequency progressive mode, only
the slice layer indicates the maximum number of block transform
coefficients contained in the particular bitstream (known as the
"priority break point"). Data partitioning is a frequency domain method
that breaks the block of 64 quantized transform coefficients into two
bitstreams. The first, higher priority bitstream contains the more
critical lower frequency coefficients and side informations (such as DC
values, motion vectors). The second, lower priority bitstream carries
higher frequency AC data.
SNR Scalability-- Similar to the point transform in JPEG, SNR
scalability is a spatial domain method where channels are coded at
identical sample rates, but with differing picture quality (achieved through
quantization step sizes). The higher priority bitstream contains base
layer data that can be added to a lower priority refinement layer to
construct a higher quality picture.
Temporal Scalability--- A temporal domain method useful in, e.g.,
stereoscopic video. The first, higher priority bitstreams codes video
at a lower frame rate, and the intermediate frames can be coded in a
second bitstream using the first bitstream reconstruction as prediction.
In stereoscopic vision, for example, the left video channel can be
prediction from the right channel.
Other scalability modes were experimented with in MPEG-2 video (such as
Frequency Scalability), but were eventually dropped in favor of methods
that demonstrated comparable or better picture quality with greater
simplicity.
39. Why MPEG-2? Wasn't MPEG-1 enough?
A. MPEG-1 was optimized for CD-ROM or applications at about 1.5
Mbit/sec. Video was strictly non-interlaced (i.e. progressive). The
international cooperation executed well enough for MPEG-1, that the committee
began to address applications at broadcast TV sample rates using the
CCIR 601 recommendation (720 samples/line by 480 lines per frame by 30
frames per second or about 15.2 million samples/sec including chroma) as
the reference.
Unfortunately, today's TV scanning pattern is interlaced. This
introduces a duality in block coding: do local redundancy areas (blocks)
exist exclusively in a field or a frame.(or a particle or wave) ? The
answer of course is that some blocks are one or the other at different
times, depending on motion activity. The additional man years of
experimentation and implementation between MPEG-1 and MPEG-2 improved
the method of block-based transform coding.
40. What did MPEG-2 add to MPEG-1 in terms of syntax/algorithms ?
A. Here is a brief summary:
Sequence layer:
More aspect ratios. A minor, yet necessary part of the syntax.
Horizontal and vertical dimensions are now required to be a multiple of
16 in frame coded pictures, and the vertical dimension must be a
multiple of 32 in field coded pictures.
4:2:2 and 4:4:4 macroblocks were added in the Next profiles.
Syntax can now signal frame sizes as large as 16383 x 16383.
Syntax signals source video type (NTSC, PAL, SECAM, MAC, component) to
help post-processing and display.
Source video color primaries (609, 170M, 240M, D65, etc.) and opto-
electronic transfer characteristics (709, 624-4M, 170M etc.) can be
indicated.
Four scaleable modes [see scalability discussion]
Picture layer:
All MPEG-2 motion vectors are specified to a half-pel sample grid.
DC precision can be user-selected as 8, 9, 10, or 11 bits.
New scalar quantization matrices may be downloaded once per picture. In High
profile, separate chrominance matrices now exist (Y and C no longer have to
share)
Concealment motion vectors were added to I-pictures in order to increase
robustness from bit errors. I pictures are the most critical and sensitive
picture in a group of pictures.
A non-linear macroblock quantization factor providing a wider dynamic
range, from 0.5 to 56, than the linear MPEG-1 (1 to 32) range. Both are
sent as a 5-bit FLC side information in the macroblock and slice
headers.
New Intra-VLC table for dct_coefficient_next (AC run-level events) that
is a better match for the histogram of Intra-coded pictures. EOB is 4
bits. The old table, dct_coef_next, are reserved for use in non-intra
pictures (P, B), although they new table can be used for Intra-coded
macroblocks in P and B pictures as well.
Alternate scanning pattern that (supposedly) improves entropy coding
performance over the original Zig-Zag scan used in H.261, JPEG, and MPEG-1.
The extra scanning pattern is geared towards interlaced video.
Syntax to signal an irregular 3:2 pulldown process (repeat_field_first flag)
Progressive and interlaced frame coding
Syntax to indicate source composite video characteristics useful in post-
processing operations. (v-axis, field sequence, sub_carrier, phase,
burst_amplitude, etc.)
Pan & scanning syntax that tells decoder how to, for example, window a
4:3 image within a wider 16:9 aspect ratio coded image. Vertical pan
offset has 1/16th pixel accuracy.
Macroblock layer:
Macroblock stuffing is now illegal in MPEG-2 (hurray!!). If stuffing is
really needed, the encoder can pad slice start codes.
Two organizations for macroblock coefficients (interlaced and progressive)
signaled by dct_type flag.
Now only one run-level escape code code (24-bits) instead of the single (20-
bits) and double escape (28-bits) in MPEG-1.
Improved mismatch control in quantization over the original oddification
method in MPEG-1. Now specifies adding or subtracting one to the 63rd
AC coefficient depending on parity of the summed coefficients. MPEG-2
mismatch control is performed on the transform coefficients, whereas in MPEG-
1, it is applied to the quantized transform coefficients.
Many additional prediction modes (16x8 MC, field MC, Dual Prime) and,
correspondingly, macroblock modes.
Overall, MPEG-2's greatest compression improvements over MPEG-1 are:
prediction modes, Intra VLC table, DC precision, non-linear macroblock
quantization. Implementation improvements: macroblock stuffing was
eliminated.
41. How do MPEG and JPEG differ?
A. The most fundamental difference is MPEG's use of block-based motion
compensated prediction (MCP)---a method falling into the general category of
temporal DPCM.
The second most fundamental difference is in the target application.
JPEG adopts a general purpose philosophy: independence from color space
(up to 255 components per frame) and quantization tables for each
component. Extended modes in JPEG include two sample precision (8 and
12 bit sample accuracy), combinations of frequency progressive, spatial
hierarchically progressive, and amplitude (point transform) progressive
scanning modes. Further color independence is made possible thanks to
downloadable Huffman tables (up to one for each component.)
Since MPEG is targeted for a set of specific applications, there is only
one color space (4:2:0 YCbCr), one sample precision (8 bits), and one
scanning mode (sequential). Luminance and chrominance share quantization
and VLC tables. MPEG adds adaptive quantization at the macroblock (16 x
16 pixel area) layer. This permits both smoother bit rate control and
more perceptually uniform quantization throughout the picture and image
sequence. However, adaptive quantization is part of the Enhanced JPEG
charter (ISO/IEC 10918-3) currently in verification stage. MPEG variable
length coding tables are non-downloadable, and are therefore optimized
for a limited range of compression ratios appropriate for the target
applications.
The local spatial decorrelation methods in MPEG and JPEG are very
similar. Picture data is block transform coded with the two-dimensional
orthanormal 8x8 DCT, with asymmetric basis vectors about time (aka DCT-
II). The resulting 63 AC transform coefficients are mapped in a zig-zag
pattern (or alternative scan pattern in MPEG-2) to statistically
increase the runs of zeros. Coefficients of the vector are then
uniformly scalar quantized, run-length coded, and finally the run-length
symbols are variable length coded using a canonical (JPEG) or modified
Huffman (MPEG) scheme. Global frame redundancy is reduced by 1-D DPCM
of the block DC coefficients, followed by quantization and variable
length entropy coding of the quantized DC coefficient.
MCP DCT ZZ
Q
Frame -> 8x8 spatial block -> 8x8 frequency block -> Zig-zag scan ->
RLC VLC
quanitzation -> run-length coding -> variable length coding.
The similarities have made it possible for the development of hard-wired
silicon that can code both standards. Even some highly microcoded
architectures employing hardwired instruction primitives or functional
blocks benefit from JPEG/MPEG similarities. There are many additional
yet minor differences. They include:
1. In addition to the 8-bit mode, DCT and quantization precision
in MPEG has a 9-bit and 12-bit mode, respectively, exclusively in non-
intra coded macroblocks. A 1-bit expansion takes place in the
macroblock difference operation.
2. Mismatch control in MPEG-1 forces quantized coefficients to
become odd values (oddification). JPEG does not employ any mismatch
mechanism.
3. JPEG run-length coding produces run-size tokens (run of zeros,
non-zero coefficient magnitude) whereas MPEG produces fully concatenated
run-level tokens that do not require magnitude differential bits.
4. DC values in MPEG-1 are limited to 8-bit precision (a constant
stepsize of 8), whereas JPEG DC precision can occupy all possible 11-
bits. MPEG-2, however, re-introduced extra DC precision critical even
at high compression ratios.
Difference between MPEG and H.261
42. How do MPEG and H.261 differ?
A. H.261, also known as Px64, was targeted for teleconferencing
applications where motion is naturally more limited. Motion vectors are
restricted to a range of +/- 15 pixel unit displacements. Prediction
accuracy is reduced since H.261 motion vectors are specified to only
integer-pel accuracy. Other quality syntactic differences include: no
B-pictures, inferior mismatch control.
43. Is H.261 the de facto teleconferencing standard?
A. Not exactly. To date, about seventy percent of the industrial
teleconferencing hardware market is controlled by PictureTel of Mass.
The second largest market controller is Compression Labs of Silicon
Valley. PictureTel hardware includes compatibility with H.261 as a
lowest common denominator, but when in communication with other
PictureTel hardware, it can switch to a mode superior at low bit rates
(less than 300kbits/sec). In fact, over 2/3 of all teleconferencing is
done at two-times switched 56 channel (~P = 2) bandwidth. ISDN is still
expensive. In each direction, video and audio are coded at an aggregate
rate of 112 kbits/sec (2*56 kbits/sec). The PictureTel proprietary
compression algorithm is acknowledged to be a combination of spatial
pyramid, lattice vector quantizer, and an unidentified entropy coding
method. Motion compensation is considerably more refined and
sophisticated than the 16x16 integer-pel block method specified in
H.261.
The Compression Labs proprietary algorithm also offers significant
improvement over H.261 when linked to other CLI hardware. Local
decorrelation is based on a DCT-VQ hybrid.
Currently, ITU-TS (International Telecommunications Union--
teleconferencing Sector), formerly CCITT, is quietly defining an
improvement to H.261 with the participation of industry vendors.
Rate control
44. What is the TM rate control and adaptive quantization technique ?
A. The Test model (MPEG-2) and Simulation Model (MPEG-1) were not, by
any stretch of the imagination, meant to epitomize state-of-the art
encoding quality. They were, however, designed to exercise the syntax,
verify proposals, and test the *relative* compression performance of
proposals in a timely manner that could be duplicated by co-
experimenters. Without simplicity, there would have been no doubt
endless debates over model interpretation. Regardless of all else, more
advanced techniques would probably trespass into proprietary territory.
The final test model for MPEG-2 is TM version 5b, aka TM version 6. The
final MPEG-1 simulation model is version 3. The MPEG-2 TM rate control
method offers a dramatic improvement over the SM method. TM adds more
accurate estimation of macroblock complexity through use of limited a
priori information. Macroblock quantization adjustments are computed on
a macroblock basis, instead of once-per-slice.
45. How does the TM work?
A. Rate control and adaptive quantization are divided into three steps:
Step One:Bit Allocation
In Complexity Estimation, the global complexity measures assign
relative weights to each picture type (I,P,B). These weights (Xi, Xp,
Xb) are reflected by the typical coded frame size of I, P, and B
pictures (see typical frame size discussion). I pictures are usually
assigned the largest weight since they have the greatest stability
factor in an image sequence. B pictures are assigned the smallest
weight since B energy do not propagate into other pictures and are usually
highly correlated with neighboring P and I pictures.
The bit target for a frame is based on the frame type, the remaining number
of bits left in the Group of Pictures (GOP) allocation, and the immediate
statistical history of previously coded pictures.
Step Two: Rate Control
Rate control attempts to adjust bit allocation if there is significant
difference between the target bits (anticipated bits) and actual coded
bits for a block of data. If the virtual buffer begins to overflow, the
macroblock quantization step size is increased, resulting in a smaller
yield of coded bits in subsequent macroblocks. Likewise, if underflow
begins, the step size is decreased. The Test Model approximates that the
target
picture has spatially uniform distribution of bits. This is a safe
approximation since spatial activity and perceived quantization noise
are almost inversely proportional. Of course, the user is free to
design a custom distribution, perhaps targeting more bits in areas that
contain text, for example.
Step Three: Adaptive Quantization
The final step modulates the macroblock quantization step size obtained in
Step 2 by a local activity measure. The activity measure itself is normalized
against the most recently coded picture of the same type (I, P, or B). The
activity for a macroblock is chosen as the minimum among the four 8x8 block
luminance variances. Choosing the minimum block is part of the concept that
a macroblock is no better than the block of highest visible distortion
(weakest link in the chain).
46. What is a good motion estimation method, then?
A. When shopping for motion vectors, the three basic characteristics
are: Search range, search pattern, and matching criteria. Search
pattern has the greatest impact on finding the best vector. Hierarchical
search patterns first find the best match between downsampled images of
the reference and target pictures and then refine the vector through
progressively higher resolutions. When compared to other fast methods,
hierarchical patterns are less likely to be confused by extremely local
distortion minimums as being a best match. Also note that subsampled search
and hierarchical search are not synonymous.
Q. Is there a limit to the length of motion vectors?
The search area is unlimited, but the reconstructed motion vectors must
not:
a. point beyond the picture boundaries (1 <= MV_x <= luminancewidth -
16) and (1 <= MV_y <= luminanceheight - 16). The - 16 is due to the
fact that the motion vector origin is the upper left hand corner of a
macroblock)
b. In Constrained Parameters MPEG-1, the motion vector is limited to a
range of [-64,+63.5] luminance samples with half-pel accuracy, and [-
128,+127.5] with integer pel accuracy. Break the constrained parameters
rules and your video sequence will not likely display on many hardware
devices.
c. In MPEG-2 Video Main Profile at Main Level, the motion vectors are
always on a half-pel co-ordinate grid, and the vertical range is
restricted to [-64, +63.5], and the horizontal limit is [-256,+255.5].
d. in MPEG-1, the syntactic limit of the motion vector is [-1024,+1023]
integer pel, horizontal and vertical.
e. in MPEG-2, the syntactic limit of the motion vector is [-2048,+2047.5]
horizontal, [-1024,+1023.5] vertical.
47. Is exhaustive search "optimal" ?
A. Definitely not in the context of block-based MCP video. Since one
motion vector represents the prediction of 256 pixels, divergent pixels
within the macroblock are misrepresented by the "global" vector. This
leads back to the general philosophy of block-based coding as an
approximation technique. In their ICASSP'93 paper, Sullivan discusses ways in
which block-based prediction schemes can solve part of this problem.
Exhaustive search may find blocks with the least distortion (displaced frame
difference) but will not produce motion vectors with the lowest entropy.
48. What are some advanced encoding methods?
Quantizer feedback: determine the dependent quantization stepsize by
modeling quantization error propagating over multiple pictures. [Uz/et
al ICASSP 93, Ortega/Vetterli/et al ICASSP 93]
Smoothness constraint placed on local activity measures. immediate blocks
outside target macroblock are considered when selecting macroblock
quantization stepsize .[Thomson/Savitier patent]
Horizontal variance: measure variance between columns of pixels in addition
to the traditional measure of variance along rows (lines) when making
field/frame macroblock prediction decision.
DFD energy: examine DFD energy/variance when making Intra/Non-intra
macroblock decision.
Activity measures: use total bits from a first-pass encoding of a picture or
macroblock as a measure of the activity. Coded bits is a more accurate
reflection of local complexity than variance. [Thomson/Savitier patent]
motion vector cost: this is true for any syntax elements, really. Signaling
a macroblock quantization factor or a large motion vector differential can
cost more than making up the difference with extra quantized DFD (prediction
error) bits. The optimum can be found with, some Lagrangian operator. In
summary, any compression system with side information, there is a optimum
point between signaling overhead (e.g. prediction) and prediction error.
Liberal Interpretations of the Forward DCT:
Borrowing from the concept that the DCT is simply a filter bank, a
technique that seems to be gaining popularity is basis vector shaping.
Usually this is combined with the quantization stage since the two are
tied closely together in a rate-distortion sense. The idea is to use
the basis vector shaping as a cheap alternative to pre-filtering by
combining the more desirable data adaptive properties of pre-filtering/
pre-processing into the transformation process... yet still reconstruct
a picture in the decoder using the standard IDCT that looks reasonably
like the source. Some more clever schemes will apply a form of windowing.
[Warning: watch out for eigenimage/basis vector orthoganality. ]
Frequency-domain enhancements:
Enhancements are applied after the DCT (and possibly quantization)stage
to the transform coefficients. This borrows from the concept: if you
don't like the (quantized) transformed results, simply reshape them into
something you do like. Suppressing isolated small amplitudes is popular.
Temporal spreading of quantization error:
This method is similar to the original intent behind color subcarrier
phase alternation by field in the NTSC, PAL, and SECAM analog TV
standards: for stationary areas, noise does not hang" in one location,
but dances about the image over time to give a more uniform effect.
Distribution makes it more difficult for the eye to "catch on" to
trouble spots (due to the latent temporal response curve of human
vision). Simple encoder models tend to do this naturally but will not
solve all situations.
Look-ahead and adaptive frame cycle structures: analyze picture activity
several pictures into the future, looking for scene changes or motion
statistics.
It is easy to spot encoders that do not employ any advanced encoding
techniques: reconstructed video usually contains ringing around edges,
color bleeding, and lots of noise.
49. Is so-and-so really MPEG compliant ?
A. At the very least, there are two areas of conformance/compliance in
MPEG: 1. Compliant bitstreams 2. compliant decoders. Technically
speaking, video bitstreams consisting entirely of I-frames (such as
those generated by Xing software) are syntactically compliant with the
MPEG specification. The I-frame sequence is simply a subset of the full
syntax. Compliant bitstreams must obey the range limits (e.g. motion
vectors limited to +/-128, frame sizes, frame rates, etc.)and syntax
rules (e.g. all slices must commence and terminate with a non-skipped
macroblock, no gaps between slices, etc.).
Decoders, however, cannot escape true conformance. For example, a
decoder that cannot decode P or B frames are *not* legal MPEG.
Likewise, full arithmetic precision must be obeyed before any decoder
can be called "MPEG compliant." The IDCT, inverse quantizer, and
motion compensated predictor must meet the specification requirements...
which are fairly rigid (e.g. no more than 1 least significant bit of
error between reference and test decoders). Real-time conformance is
more complicated to measure than arithmetic precision, but it is
reasonable to expect that decoders that skip frames on reasonable
bitstreams are not likely to be considered compliant.
Artifacts
50. What are the tell-tale MPEG artifacts?
A. If the encoder did its job properly, and the user specified a proper
balance between sample rate and bitrate, there shouldn't be any visible
artifacts. However, in sub-optimal systems, you can look for:
Gibbs phenomenon/Ringing/Aliasing (too few AC bits, not enough
pre-processing)
Blockiness (not considering your neighbors before quantizing)
Posterization (too few DC bits)
Checkerboards (DCT eigenimages as a result of too few AC coefficients)
Colorbleeding (not considering color in encoder cost model, not
subtracting color at edges of objects, etc.)
51. Where are the weak points of MPEG video ?
A.
Texture patterns (rapidly alternating lines)
sharp edges (especially text)
[installment 3]
52. What are some myths about MPEG?
A. There are a few major myths that I am aware of:
1. Block displacements: macroblock predictions are formed out of
arbitrary 16x16 (or 16x8/16x16 in MPEG-2) areas from previously
reconstructed pictures. Many people believe that the prediction
macroblocks have boundaries that fall on interchange boundaries (pixel
0, 15, 31, 53... line 0, 15, 31, 53... etc.). In fact, motion vectors
represent relative translations with respect to the target
reconstruction macroblock coordinates. The motion vectors can point to
half pixel coordinates, requiring that the prediction macroblock to be
formed via bi-linear interpolation of pixels.
2. Displaced frame (macroblock) difference construction: the prediction
error formed as the difference between the prediction macroblock and
source macroblock is coded much like an Intra macroblock. The
prediction may come from different locations (as in bi-directional
prediction--or in MPEG-2--16x8, field-in-frame, and Dual Prime), but the
DFD is always coded as a 16x16 unit.
3. Compression ratios
You hear 200:1 and 100:1 in the media. Utter rubbish. The true range
is between 16:1 and 40:1. Spreading misinformation about compression
ratios in public will catch the attention of the infamous MPEG Police.
They say mild-mannered Michael Barnsley will snap, without warning, into
violent rage if he doesn't get the upper bunk bed.
4. Picture coding types all consist of the same macroblocks
Macroblocks within I pictures are strictly intra-coded. Macroblocks
within P pictures can be either predicted or intra-coded, and B pictures
they can be bi-directional, forward, backward, or intra. Additional
macroblock modes switches include: predicted with no motion
compensation, modified macroblock quantization, coding of prediction error or
not. The switches are concatenated into the macroblock_type side information
and variable length coded in the macroblock header.
53. What is the color space of MPEG?
MPEG strictly specifies the YCbCr color space, not YUV or YIQ or YPbPr
or YDrDb or any other color difference variations. Regardless of any
bitstream parameters, MPEG-1 and MPEG-2 Video Main Profile specify 4:2:0
chroma ratio, where the color difference channels (Cb, Cr) have half the
resolution or sample grid density in both the horizontal and vertical
direction
with respect to luminance.
MPEG-2 High Profile includes an option for 4:2:2 and 4:4:4 coding.
Applications
for this are likely to be broadcasting and contribution equipment.
54. Don't you mean 4:1:1 ?
A. No, here is a table of ratios:
CCIR 601 (60 Hz) image Chroma sub-sampling factors
format Y Cb, Cr Vertical Horizontal
----- --------- ---------- -------- ----------
4:4:4 720 x 480 720 x 480 none none
4:2:2 720 x 480 360 x 480 none 2:1
4:2:0 720 x 480 360 x 240 2:1 2:1
4:1:1 720 x 480 720 x 120 none 4:1
4:1:0 720 x 480 180 x 120 4:1 4:1
3:2:2, 3:1:1, and 3:1:0 are less common variations.
55. Why did MPEG choose 4:2:0 ? Isn't 4:2:2 the standard for TV?
A. At least three reasons I can think of:
1. 4:2:0 picture memory requirements are 33% less than the size of 4:2:2
pictures.
MPEG-1 decoder are able to snugly fit all 3 SIF pictures (1 reconstruction &
display, 2 prediction) into 512 KBytes of buffer space. CCIR 601 is a
tighter fit into 2 Mbytes.
2. The subjective difference between 4:2:0 and 4:2:2 is minimal, when
considering consumer display equipment and distribution compression ratios.
3. Vertical decimation increases compression efficiency by reducing syntax
overhead posed in an 8 block (4:2:0) macroblock structure.
4. You re compressing the hell out of the video signal, so what possible
difference can the 0:0:2 high-pass make?
Interlacing and the 62 microsecond gap between successively scanned lines
introduces some discontinuities, but most of this can be alleviated through
pre-processing.
56. What is the precision of MPEG samples?
A. By definition, MPEG samples have no more and no less than 8-bits uniform
sample precision (256 quantization levels). For luminance (which is
unsigned) data, black corresponds to level 0, white is level 255. However, in
CCIR recommendation 601 chromaticy, levels 0 through 14 and 236 through 255
are reserved for blanking signal excursions. MPEG currently has no such
clipped excursion restrictions, although decoder might take care to insure
active samples do not exceed these limits. With three color components per
pixel, the total combination is roughly 16.8 million colors (i.e. 24-bits).
57. What is all the fuss with cositing of chroma components?
A. It is moderately important to properly co-site chroma samples,
otherwise a sort of chroma shifting effect (exhibited as a halo) may result
when the reconstructed video is displayed. In MPEG-1 video, the chroma
samples are exactly centered between the 4 luminance samples (Fig 1.) To
maintain compatibility with the CCIR 601 horizontal chroma locations and
simplify implementation (eliminate need for phase shift), MPEG-2 chroma
samples are arranged as per Fig.2.
Y Y Y Y Y Y Y Y YC Y YC Y
C C C C
Y Y X Y Y Y Y Y YC Y YC Y
Y Y Y Y Y Y Y Y YC Y YC Y
C C C C
Y Y Y Y Y Y Y Y YC Y YC Y
Fig.1 MPEG-1 Fig.2 MPEG-2 Fig.3 MPEG-2 and
4:2:0 organization 4:2:0 organization CCIR Rec. 601
4:2:2 organization
MPEG for the data compression expert
58. How would you explain MPEG to the data compression expert?
A. MPEG video is a block-based video scheme.
59. How does MPEG video really compare to TV, VHS, laserdisc ?
A. VHS picture quality can be achieved for source film video at about 1
million bits per second (with proprietary encoding methods). It is very
difficult to objectively compare MPEG to VHS. The response curve of
VHS places -3 dB at around 2 MHz of analog luminance bandwidth
(equivalent to 200 samples/line). VHS chroma is considerably less dense
in the horizontal direction than MPEG source video (compare 80
samples/line to 176!). From a sampling density perspective, VHS is
superior only in the vertical direction (480 luminance lines compared to
240)...
but when taking into account (supposedly such things as) interfield magnetic
tape crosstalk and the TV monitor Kell factor, the perceptual vertical
advantage is not all that significant. VHS is prone to such inconveniences
as timing errors (an annoyance addressed by time base correctors), whereas
digital video is fully discretized. Pre-recorded VHS is typically recorded at
very high duplication speeds (5 to 15 times real time playback speed),
opening up additional avenues for artifacts. In gist, MPEG-1 at its nominal
parameters can match VHS's sexy low-pass-filtered look.
With careful coding schemes, broadcast NTSC quality can be approximated at
about 3 Mbit/sec, and PAL quality at about 4 Mbit/sec. Of course, sports
sequences with complex spatial-temporal activity should be treated with bit
rates more like 5 and 6 Mbit/sec, respectively. Laserdisc is a tough one to
compare. Laserdisc's are encoded with composite video (NTSC or PAL).
Manufacturers of laser disc players make claims of up to 425 TVL (or 567
samples/line) response. Thus it could be said the laserdisc has a 567 x 480 x
30 Hz "potential resolution". The carrier-to-noise ratio is typically better
than 48 dB. Timing is excellent. Yet some of the clean characteristics of
laserdisc can be achieved with MPEG-1 at 1.15 Mbit/sec (SIF rates),
especially for those areas of medium detail (low spatial activity) in the
presence of uniform motion. This may be why some people say MPEG-1 video at
1.15 Mbit/sec looks almost as good as Laserdisc or Super VHS at times.
60. What are the typical MPEG-2 bitrates and picture quality?
Picture type
I P B Average
MPEG-1 SIF
@ 1.15 Mbit/sec 150,000 50,000 20,000
38,000
MPEG-2 601 400,000 200,000 80,000
130,000
@ 4.00 Mbit/sec
Note: parameters assume Test Model for encoding, I frame distance of 15 (N =
15), and a P frame distance of 3 (M = 3).
Of course, among differing source material, scene changes, and use of
advanced encoder models... these numbers can be significantly different.
61. At what bitrates is MPEG-2 video optimal?
A. The Test subgroup has defined a few examples:
"Sweet spot" sampling dimensions and bit rates for MPEG-2:
Dimensions Coded rate Comments
------------- ---------- ----------------------------------------
---
352x480x24 Hz 2 Mbit/sec Half horizontal 601. Looks almost NTSC
(progressive) broadcast quality, and is a good
(better)
substitute for VHS. Intended for film src.
544x480x30 Hz 4 Mbit/sec PAL broadcast quality (nearly full
capture
(interlaced) of 5.4 MHz luminance carrier). Also
4:3 image dimensions windowed within 720
sample/line 16:9 aspect ratio via pan&scan.
704x480x30 Hz 6 Mbit/sec Full CCIR 601 sampling dimensions.
(interlaced)
[these numbers subject to change at whim of MPEG Test subgroup]
62. Why does film perform so well with MPEG ?
A. Several reasons, really:
1) The frame rate is 24 Hz (instead of 30 Hz) which is a savings of
some 20%.
2) the film source video is inherently progressive. Hence no fussy
interlaced spectral frequencies.
3) the pre-digital source was severely oversampled (compare 352 x 240
SIF to 35 millimeter film at, say, 3000 x 2000 samples). This can
result in a very high quality signal, whereas most video cameras
do
not oversample, especially in the vertical direction.
4) Finally, the spatial and temporal modulation transfer function
(MTF)
characteristics (motion blur, etc) of film are more amenable to
the transform and quantization methods of MPEG.
63. What is the best compression ratio for MPEG ?
A. The MPEG sweet spot is about 1.2 bits/pel Intra and .35 bits/pel
inter. Experimentation has shown that intra frame coding with the
familiar DCT-Quantization-Huffman hybrid algorithm achieves optimal
performance at about an average of 1.2 bits/sample or about 6:1
compression ratio. Below this point, artifacts become noticeable.
64. Can MPEG be used to code still frames?
A. Yes. There are, of course, advantages and disadvantages to using
MPEG over JPEG:
Disadvantages:
1. MPEG has only one color space
2. MPEG-1 and MPEG-2 Main Profile luma and chroma share quanitzation
and VLC tables
3. MPEG-1 is syntactically limited to 4k x 4k images, and 16k x 16k for
MPEG-2.
Advantages:
1. MPEG possesses adaptive quantization
2. With its limited still image syntax, MPEG averts any temptation to use
unnecessary, expensive, and academic encoding methods that have little
impact on the overall picture quality (you know who you are).
Philips' CD-I spec. has a requirement for a MPEG still frame mode, with
double SIF image resolution. This is technically feasible mostly thanks to
the fact that only one picture buffer is needed to decode a still image
instead of three buffers.
65. Is there an MPEG file format?
A. Not exactly. The necessary signal elements that indicate image size,
picture rate, aspect ratio, etc. are already contained within the sequence
layer of the MPEG video stream. The Whitebook format for Karoke and CD-I
movies specify a range of (time-division) multiplexing strategies for audio
and video bitstreams. A directory format listing scenes and their locations
on the disc is associated with the White Book specification.
66. What are some pre-processing enhancements ?
Adaptive de-interlacing:
This method maps interlaced video from a higher sampling rate (e.g 720 x 480)
into a lower rate, progressive format (352 x 240). The most basic algorithm
measures the correlation between two immediate macroblock fields, and if the
correlation is high enough, uses an average of both fields to form a frame
macroblock. Otherwise, a field area from one field (usually of the same
parity) is selected. More clever algorithms are much more complex than this,
and may involve median filtering, and multirate/multidimensional tools.
Pre-anti-aliasing and Pre-blockiness reduction:
A common method in still image coding is to pre-smooth the image before
encoding. For example, if pre-analysis of a frame indicates that serious
artifacts will arise if the picture were to be coded in the current condition
(i.e. below the sweet spot), a pre-anti-aliasing filter can be applied. This
can be as simple as having a smoothing severity proportional to the image
activity. The pre-filter can be global (same smoothing factor for whole
image or sequence) or locally adaptive. More complex methods will again use
multirate/multidimensional methods.
One straightforward concept from multidimensional/multirate e-processing is
to apply source video whose resolution (sampling density) is greater than
the target source and reconstruction sample rates. This follows the basic
principles of oversampling, as found in A/D converters.
These filters emphasize the fact that most information content is contained
in the lower harmonics of a picture anyway. VHS is hardly considered to be a
sharp cut-off medium, tragically implying that "320 x 480 potential" of
VHS is never truly realized.
67. Why use these "advanced" pre-filtering techniques?
A. Think of the DCT and quantizer as an A/D converter. Think of the DCT/Q
pre-filter as the required anti-alias prefilter found before every A/D. The
big difference of course is that the DCT quantizer assigns a varying number
of bits per transform coefficient. Judging on the normalized activity
measured in the pre-analysis stage of video encoding (assuming you even have
a pre-analysis stage), and the target buffer size status, you have a fairly
good idea of how many bits can be spared for the target macroblock, for
example.
Other pre-filtering techniques mostly take into account: texture patterns,
masking, edges, and motion activity. Many additional advanced techniques can
be applied at different immediate layers of video encoding (picture, slice,
macroblock, block, etc.).
68. What about post-processing enhancements?
Some research has been carried out in this area. Non-linear interpolation
methods have been published by Wu and Gersho (e.g. ICASSP 93), convex hull
projections for MAP (Severinson, ICASSP 93), and others. Post-processing
unfortunately defies the spirit of MPEG conformance. Decoders should produce
similar reconstructions. Enhancements should ideally be done during the pre-
processing and encoding stages.
69. Can motion vectors be used to measure object velocity?
A. Motion vector information cannot be reliably used as a means of
determining object velocity unless the encoder model specifically set
out to do so. First, encoder models that optimize picture quality generate
vectors that typically minimize prediction error and, consequently,
the vectors often do not represent true object translation. Standards
converters that resample one frame rate to another (as in NTSC to PAL)
use different methods (motion vector field estimation, edge detection, et
al) that are
not concerned with optimizing ratios such as SNR vs bitrate. Secondly, motion
vectors
are not transmitted for all macroblocks anyway.
70. How do you code interlaced video with MPEG-1 syntax?
A. Two methods can be applied to interlaced video that maintain
syntactic compatibility with MPEG-1 (which was originally designed for
progressive frames only). In the field concatenation method, the
encoder model can carefully construct predictions and prediction errors
that realize good compression but maintain field integrity (distinction
between adjacent fields of opposite parity). Some pre-processing
techniques can also be applied to the interlaced source video that
would, e.g., lessen sharp vertical frequencies.
This technique is not efficient of course. On the other hand, if the
original source was progressive (e.g. film), then it is more trivial to
convert the interlaced source to a progressive format before encoding.
(MPEG-2 would then only offer superior performance through greater DC
block precision, non-linear mquant, intra VLC, etc.) Reconstructed
frames are re-interlaced in the decoder Display process.
The second syntactically compatible method codes fields as separate pictures.
This approach has been acknowledged not to work as well.
71. Is MPEG patented?
A. Yes and no. Many encoding methods are patented. Approximately 11
blocking patents, that is, patents that are general enough to be unavoidable
in any implementation have been recently identified.
A patent pool is being formed within MPEG where a single royalty fee would be
split among the 31 patent-holding companies.
72. How many cable box alliances are there?
A. Many. To start with:
Scientific Atlanta (SA), Kaledia, and Motorola:
SA will build the box, Motorola the chips, and Kaleida the
O/S and user interface (using ScriptX of course).
Silicon Graphics (SGI), Scientific Atlanta, and Toshiba
For the Time Warner's Orlando trial, SGI will provide the
RISC (MIPS R4000) and software, SA will do the box again,
and Toshiba will provide the chips.
General Instruments (GI) and Microsoft:
GI will make the box and Intel will supply the special low-cost
386SL processor on which a 1MB flash EPROM executable core
of Microsoft windows and DOS will run. Microsoft will develop the
user interface.
Hewlett Packard (HP):
HP will manufacture and/or design low cost, open architecture set-top
decoder boxes (not a part of the Eon wireless deal). The CPU will
explicitly not use a 80x68 based processor.
CLI and Philips:
Compression Labs will provide the encoder technology and Philips
will provide the decoder techology for an ADSL system whose
transport structure will be put together by Broadband Technologies.
["These alliances subject to change at the whim of PR departments
and market forces."]
73. Will there be an MPEG video tape format?
A. Not exactly. A consortium of international companies are co-
developing a consumer digital video 6 millimeter wide, metal particle
tape format. Due to the initial high cost of MPEG encoders, a JPEG-like
compression method will be used for inexpensive encoding of typical
consumer source video (broadcast PAL, NTSC). The natural consequence of
still image methods is less efficient use of bandwidth: 25 Mbit/sec for
the same subjective real-time playback quality achieved at 6 Mbit/sec
possible with MPEG-2. A second bit rate mode, 50 Mbit/sec, is
designated for HDTV.
Pre-coded digital video from, e.g., broadcast sources will be directly
recorded to tape and "passed-through" as a coded bitstream to the video
decompression box upon tape playback. Assuming if linear tape speed is
to be proportional to bit rate, the recording time of a pre-compressed
MPEG-2 program at the upper limit of 5 Mbit/sec for broadcast quality
video, the recording time would be over 20 hours. Channel coding
schemes (error correction, convolution coding, etc.), however, will
most likely be optimized for the tape medium and therefore may differ
from the channel methods for cable, terrestrial, and satellite. (A
Zenith-Goldstar S-VHS based experiment did, however, directly record the
4-VSB broadcast baseband signal of the old Zenith/AT&T HDTV proposal).
More specs: (Summarized from EE Times July 5, 1993 article)
tape width: 6.35 mm
Audio: two channel 48 KHz 16-bit audio, or 4 channel at 32 KHz at 12-bit
Tape format: metal evaporated tape, 13.5 microns thick
Cassette dimensions: (millimeters) Recording times:
Size Width Height Depth 525/625 (25Mb/sec) HDTV (50 Mb/s)
-------- ----- ------ ----- ------------------ --------------
Standard 125 78 14.6 4h30min 2h15min
Small 66 48 12.2 1 hour 30min
Linear tape speeds: 18.812 mm/s (60Hz), 18.831 mm/s (50 Hz)
Video compression: DCT based
Participants: Matsushita, Sony, Philips, Thomson, Hitachi, Mitsubishi,
Sanyo, Sharp, Toshiba, JVC.
MPEG in everyday life
74. Where will be see MPEG in everyday life?
A. Just about wherever you see video today.
DBS (Direct Broadcast Satellite)
The Hughes/USSB DBS service will use MPEG-2 video and audio. Thomson
has exclusive rights to manufacture the decoding boxes for the first 18
months of operation. Hughes/USSB DBS will begin its U.S. service in
April 1994. Two satellites at 101 degrees West will share the power
requirements of 120 Watts per 27 MHz transponder over a total of 32
transponders. Multi source channel rate control methods will be
employed to optimally allocate bits between several programs normalized
to one 22 Mbit/sec data carrier. Bit allocation adapts to instantaneous co-
channel
spatial and co-channel temporal activity. An average of 150 channels are
planned with the addition of a second set of satellites augmenting the power
level of each transponder to 240 Watts. The coded throughput of each
transponder will increase to 30 Mbit/sec.
CATV (Cable Television)
Despite conflicting options, the cable industry has more or less
settled on MPEG-2 video. Audio is less than settled. For example,
General Instruments (the largest U.S. consumer cable set-top box
manufacturer) have announced the planned exclusive use of Dolby AC-3.
The General Instruments DigiCipher I video syntax is similar to MPEG-2
syntax, but employs smaller macroblock predictions and no B-frames. The
DigiCipher II specification will include modes to support both the GI
and full MPEG-2 Video Main Profile syntax. Digicipher-I services such
as HBO will upgrade to DigiCipher II in 1994.
HDTV
The U.S. Grand Alliance, a consortium of companies that formerly competed
to win the U.S. terrestrial HDTV standard, have already agreed to
use the MPEG-2 Video and Systems syntax---including B-pictures. Both
interlaced(1920 x 1080 x 30 Hz) and progressive (1280 x 720 x 60 Hz)
modes will be supported. The Alliance has also settled upon a modulation
method (VSB) convolution coding (Viterbi), and error correction (Reed-
Soloman) specification.
In September 1993, the consortium of 85 European companies signed an
agreement to fund a project known Digital Video Broadcasting (DVB) which
will develop a standard for cable and terrestrial transmission by the
end of 1994. The scheme will use MPEG-2. This consortium has put the
final nail in the coffin of the D-MAC scheme for gradual migration
towards an all-digital, HDTV consumer transmission standard. The only
remaining analog or digital-analog hybrid system left in the world is
NHK's MUSE (which will probably be axed in a few years as soon as it appears
to be politically secure thing to do).
75. What is the best compression ratio for MPEG ?
A. The MPEG sweet spot is about 1.2 bits/pel Intra and .35 bits/pel
inter. Experimentation has shown that intra frame coding with the
familiar DCT-Quantization-Entropy hybrid algorithm achieves optimal
performance at about an average of 1.2 bits/sample or about 6:1
compression ratio. Below this point, artifacts become noticeable.
76. Is there a MPEG CD-ROM format?
A. Yes, a consortium of international companies (Matsushita, Philips,
Sony, JVC, et al) have agreed upon a specification for MPEG video and
audio. 2 hour long movies are stored on two 650 MByte compact discs. The
video
rate is 1.15 Mbit/sec, the audio rate is either 128 kbit/sec or 192 kbit/sec
Layer I or Layer II.(this seems to contradict the Philips 224 kbit/s audio
spec?). Although the Video, Systems, and Audio syntax are identical, the CD-I
movie format and the White Book format are not compatible.
Researchers are busy experimenting with denser and faster rate CD
formats, perhaps using green or blue laser wavelengths. One demonstration
stretched the pit and track density to its limits, improving areal density by
almost 2 fold.