1Ppmtompeg User Manual(0) Ppmtompeg User Manual(0)
2
6 ppmtompeg - encode an MPEG-1 bitstream
7
10 ppmtompeg [options] parameter-file
11
14 This program is part of Netpbm(1).
15 ppmtompeg produces an MPEG-1 video stream. MPEG-1 is the first great
17 video compression method, and is what is used in Video CDs (VCD). ppm‐
18 tompeg originated in the year 1995. DVD uses a more advanced method,
19 MPEG-2. There is an even newer method called MPEG-4 which is also
20 called Divx. I don't know where one finds that used.
21 There's technically a difference between a compression method for video
23 and an actual file (stream) format for a movie, and I don't know if it
24 can be validly said that the format of the stream ppmtompeg produces is
25 MPEG-1.
26 Mencoder from the Mplayer package ⟨http://www.mplayerhq.hu⟩ is proba‐
28 bly superior for most video format generation needs, if for no other
29 reason than that it is more popular.
30 The programming library PM2V ⟨http://pm2v.free.fr⟩ generates MPEG-2
32 streams.
33 Use Mplayer ⟨http://www.mplayerhq.hu⟩ (not part of Netpbm) to do the
35 reverse conversion: to create a series of PNM files from an MPEG
36 stream.
37 param_file is a parameter file which includes a list of input files and
39 other parameters. The file is described in detail below.
40 To understand this program, you need to understand something about the
42 complex MPEG-1 format. One source of information about this standard
43 format is the section Introduction to MPEG in the Compression FAQ
44 ⟨http://www.faqs.org/faqs/compression-faq⟩ .
45
48 The -gop, -combine_gops, -frames, and -combine_frames options are all
49 mutually exclusive.
50 -stat stat_file
54 This option causes ppmtompeg to append the statistics that it
55 write to Standard Output to the file stat_file as well. The
56 statistics use the following abbreviations: bits per block
57 (bpb), bits per frame (bpf), seconds per frame (spf), and bits
58 per second (bps).
59 These statistics include how many I, P, and B frames there were,
61 and information about compression and quality.
62 -quiet num_seconds
66 causes ppmtompeg not to report remaining time more often than
67 every num_seconds seconds (unless the time estimate rises, which
68 will happen near the beginning of the run). A negative value
69 tells ppmtompeg not to report at all. 0 is the default (reports
70 once after each frame). Note that the time remaining is an
71 estimate and does not take into account time to read in frames.
72 -realquiet
75 causes ppmtompeg to run silently, with the only screen output
76 being errors. Particularly useful when reading input from
77 stdin.
78 -no_frame_summary
82 This option prevents ppmtompeg from printing a summary line for
83 each frame
84 -float_dct
87 forces ppmtompeg to use a more accurate, yet more computation‐
88 ally expensive version of the DCT.
89 -gop gop_num
92 causes ppmtompeg to encode only the numbered GOP (first GOP is
93 0). The parameter file is the same as for normal usage. The
94 output file will be the normal output file with the suffix
95 .gop.gop_num. ppmtompeg does not output any sequence informa‐
96 tion.
97 -combine_gops
100 causes ppmtompeg simply to combine some GOP files into a single
101 MPEG output stream. ppmtompeg inserts a sequence header and
102 trailer. In this case, the parameter file needs only to contain
103 the SIZE value, an output file, and perhaps a list of input GOP
104 files (see below).
105 If you don't supply a list of input GOP files is used, then ppm‐
107 tompeg assumes you're using the same parameter file you used
108 when you created the input (with the -gop option) and calculates
109 the corresponding gop filenames itself. If this is not the
110 case, you can specify input GOP files in the same manner as nor‐
111 mal input files -- except instead of using INPUT_DIR, INPUT, and
112 END_INPUT, use GOP_INPUT_DIR, GOP_INPUT, and GOP_END_INPUT. If
113 no input GOP files are specified, then the default is to use the
114 output file name with suffix .gop.gop_num, with gop_num starting
115 from 0, as the input files.
116 Thus, unless you're mixing and matching GOP files from different
118 sources, you can simply use the same parameter file for creating
119 the GOP files (-gop) and for later turning them into an MPEG
120 stream (-combine_gops).
121 -frames first_frame last_frame
125 This option causes ppmtompeg to encode only the frames numbered
126 first_frame to last_frame, inclusive. The parameter file is the
127 same as for normal usage. The output will be placed in separate
128 files, one per frame, with the file names being the normal out‐
129 put file name with the suffix .frame.frame_num. No GOP header
130 information is output. (Thus, the parameter file need not
131 include the GOP_SIZE value)
132 Use ppmtompeg -combine_frames to combine these frames later into
134 an MPEG stream.
135 -combine_frames
139 This option causes ppmtompeg simply to combine some individual
140 MPEG frames (such as you might have created with an earlier run
141 of ppmtompeg -frames) into a single MPEG stream. Sequence and
142 GOP headers are inserted appropriately. In this case, the
143 parameter file needs to contain only the SIZE value, the
144 GOP_SIZE value, an output file, and perhaps a list of frame
145 files (see below).
146 The parameter file may specify input frame files in the same
148 manner as normal input files -- except instead of using
149 INPUT_DIR, INPUT, and END_INPUT, use FRAME_INPUT_DIR,
150 FRAME_INPUT, and FRAME_END_INPUT. If no input frame files are
151 specified, then the default is to use the output file name with
152 suffix .frame.frame_num, with frame_num starting from 0, as the
153 input files.
154 -nice This option causes ppmtompeg to run any remote processes
159 "nicely," i.e. at low priority. (This is relevant only if you
160 are running ppmtompeg in parallel mode. Otherwise, there are no
161 remote processes). See 'man nice.'
162 -max_machines num_machines
165 This option causes ppmtompeg to use no more than num_machines
166 machines as slaves for use in parallel encoding.
167 -snr This option causes ppmtompeg to include the signal-to-noise
170 ratio in the reported statistics. Prints SNR (Y U V) and peak
171 SNR (Y U V) for each frame. In summary, prints averages of
172 luminance only (Y). SNR is defined as 10*log(variance of origi‐
173 nal/variance of error). Peak SNR is defined as
174 20*log(255/RMSE). Note that ppmtompeg runs a little slower when
175 you use this option.
176 -mse This option causes ppmtompeg to report the mean squared error
179 per block. It also automatically reports the quality of the
180 images, so there is no need to specify -snr then.
181 -bit_rate_info rate_file
184 This option makes ppmtompeg write bit rate information into the
185 file rate_file. Bit rate information is bits per frame, and
186 also bits per I-frame-to-I-frame.
187 -mv_histogram
190 This option causes ppmtompeg to print a histogram of the motion
191 vectors as part of statistics. There are three histograms --
192 one for P frame, one for forward B frame, and one for backward B
193 frame motion vectors.
194 The output is in the form of a matrix, each entry corresponding
196 to one motion vector in the search window. The center of the
197 matrix represents (0,0) motion vectors.
198 -debug_sockets
201 This option causes ppmtompeg to print to Standard Output mes‐
202 sages that narrate the communication between the machines when
203 you run ppmtompeg in parallel mode ⟨#parallel⟩ .
204 -debug_machines
207 This option causes ppmtompeg to print to Standard Output mes‐
208 sages that narrate the progress of the conversion on the various
209 machines when you run ppmtompeg in parallel mode ⟨#parallel⟩ .
210
215 The parameter file must contain the following lines (except when using
216 the -combine_gops or -combine_frames options):
217 PATTERN pattern
222 This statement specifies the pattern (sequence) of I frames, P
223 frames, and B frames. pattern is just a sequence of the letters
224 I, P, and B with nothing between. Example:
225 PATTERN IBBPBBPBBPBBPBB
227 See I Frames, P Frames, B Frames ⟨#ipb⟩ .
229 OUTPUT output file
232 This names the file where the output MPEG stream goes.
233 INPUT_DIR directory
236 This statement tells where the input images (frames) come from.
237 If each frame is in a separate file, directory is the directory
238 where they all are. You may use . to refer to the current
239 directory. A null directory refers to the root directory of the
240 system file tree.
241 To have ppmtompeg read all the frames serially from Standard
243 Input, specify
244 INPUT_DIR stdin
245 INPUT This line must be followed by a list of the input files (in dis‐
248 play order) and then the line END_INPUT.
249 There are three types of lines between INPUT and END_INPUT.
251 First, a line may simply be the name of an input file. Second,
252 the line may be of the form single_star_expr [x-y]. sin‐
253 gle_star_expr can have a single * in it. It is replaced by all
254 the numbers between x and y inclusive. So, for example, the
255 line tennis*.ppm [12-15] refers to the files tennis12.ppm, ten‐
256 nis13.ppm, tennis14.ppm, tennis15.ppm.
257 Uniform zero-padding occurs, as well. For example, the line
259 football.*.ppm [001-130] refers to the files football.001.ppm,
260 football.002.ppm, ..., football.009.ppm, football.010.ppm, ...,
261 football.130.ppm.
262 The third type of line is: single_star_expr [x-y+s], where the
264 line is treated exactly as above, except that we skip by s.
265 Thus, the line football.*.ppm [001-130+4] refers to the files
266 football.001.ppm, football.005.ppm, football.009.ppm, foot‐
267 ball.013.ppm, etc.
268 Furthermore, a line may specify a shell command to execute to
270 generate lines to be interpreted as described above, as if those
271 lines were in the parameter file instead. Use back ticks, like
272 in the Bourne Shell, like this:
273 `cat myfilelist`
275 If input is from Standard Input (per the INPUT_DIR statement),
277 ppmtompeg ignores the INPUT/END_INPUT block, but it still must
278 be present.
279 BASE_FILE_FORMAT {PPM | PNM | YUV |
282 JPEG | JMOVIE} ppmtompeg must convert all input files to
283 one of the following formats as a first step of processing: PNM,
284 YUV, JPEG(v4), or JMOVIE. (The conversion may be trivial if
285 your input files are already in one of these formats). This
286 line specifies which of the four formats. PPM is actually a
287 subset of PNM. The separate specification is allowed for back‐
288 ward compatibility. Use PNM instead of PPM in new applications.
289 INPUT_CONVERT conversion_command
292 You must specify how to convert a file to the base file format.
293 If no conversion is necessary, then you would just say:
294 INPUT_CONVERT *
296 Otherwise, conversion_command is a shell command that causes an
298 image in the format your specified with BASE_FILE_FORMAT to be
299 written to Standard Output. ppmtompeg executes the command once
300 for each line between INPUT and END_INPUT (which is normally,
301 but not necessarily, a file name). In the conversion command,
302 ppmtompeg replaces each '*' with the contents of that line.
303 If you had a bunch of gif files, you might say:
305 INPUT_CONVERT giftopnm *
306 If you have a bunch of separate a.Y, a.U, and a.V files
308 (where
309 the U and V have already been subsampled), then you might
310 say:
311 INPUT_CONVERT cat *.Y *.U *.V
313 Input conversion is not allowed with input from stdin, so use
315 INPUT_CONVERT *
317 as described above.
319 SIZE widthxheight
322 width and height are the width and height of each frame in pix‐
324 els.
325 When ppmtompeg can get this information from the input image
327 files, it ignores the SIZE parameter and you may omit it.
328 When the image files are in YUV format, the files don't contain
330 dimension information, so SIZE is required.
331 When ppmtompeg is running in parallel mode, not all of the pro‐
333 cesses in the network have access to the image files, so SIZE is
334 required and must give the same dimensions as the input image
335 files.
336 YUV_SIZE widthxheight
339 This is an obsolete synonym of SIZE.
340 YUV_FORMAT {ABEKAS | PHILLIPS | UCB |
343 EYUV | pattern} This is meaningful only
344 when BASE_FILE_FORMAT specifies YUV format, and then it is
345 required. It specifies the sub-format of the YUV class.
346 GOP_SIZE n
350 n is the number of frames in a Group of Pictures. Except that
351 because a GOP must start with an I frame, ppmtompeg makes a GOP
352 as much longer than n as it has to to make the next GOP start
353 with an I frame.
354 Normally, it makes sense to make your GOP size a multiple of
356 your pattern length (the latter is determined by the PATTERN
357 parameter file statement).
358 See Group Of Pictures ⟨#gop⟩ .
360 SLICES_PER_FRAME n
363 n is roughly the number of slices per frame. Note, at least one
364 MPEG player may complain if slices do not start at the left side
365 of an image. To ensure this does not happen, make sure the num‐
366 ber of rows is divisible by SLICES_PER_FRAME.
367 PIXEL {FULL | HALF}
370 use half-pixel motion vectors, or just full-pixel ones It is
371 usually important that you use half-pixel motion vectors,
372 because it results in both better quality and better compres‐
373 sion.
374 RANGE n
378 Use a search range of n pixels in each of the four directions
379 from a subject pixel. (So the search window is a square n*2
380 pixels on a side).
381 PSEARCH_ALG {EXHAUSTIVE | TWOLEVEL |
384 SUBSAMPLE | LOGARITHMIC} This statement tells ppmtompeg
385 what kind of search
386 technique (algorithm) to use for P frames. You select the
387 desired
388 combination of speed and compression. EXHAUSTIVE gives the
389 best compression, but LOGARITHMIC is the fastest.
390 TWOLEVEL is an exhaustive full-pixel search, followed by a
391 local half- pixel search around the best full-pixel vector
392 (the
393 PIXEL option is ignored for this search technique).
394 BSEARCH_ALG {SIMPLE | CROSS2 | EXHAUSTIVE}
397 This statement tells ppmtompeg what kind of search
398 technique (algorithm) to use for B frames. SIMPLE means
399 find best forward and backward vectors, then interpolate.
400 CROSS2 means find those two vectors, then see what backward
401 vector best matches the best forward vector, and vice versa.
402 EXHAUSTIVE does an n-squared search and is
403 extremely slow in relation to the others (CROSS2
404 is about half as fast as SIMPLE).
405 IQSCALE n
408 Use n as the qscale for I frames.
409 See Qscale ⟨#qscale⟩ .
410 PQSCALE n
413 Use n as the qscale for P frames.
414 See Qscale ⟨#qscale⟩ .
415 BQSCALE n
418 Use n as the qscale for B frames.
419 See Qscale ⟨#qscale⟩ .
420 REFERENCE_FRAME {ORIGINAL | DECODED}
423 This statement determines whether ppmtompeg uses the original
424 images or the decoded images when computing motion vectors.
425 Using decoded images is more accurate and should increase the
426 playback quality of the output, but it makes the encoding take
427 longer and seems to give worse compression. It also causes some
428 complications with parallel encoding. (see the section on paral‐
429 lel encoding). One thing you can do as a trade-off is select
430 ORIGINAL here, and lower the qscale (see QSCALE if the quality
431 is not good enough.
432 Original or Decoded? (Normalized)
434 ────────────────────────────────────────────────────────────────────
436 Reference Compression Speed Quality I Quality P Quality B
437 Decoded 1000 1000 1000 969 919
438 Original 885 1373 1000 912 884
439 The following lines are optional:
446 FORCE_ENCODE_LAST_FRAME
451 This statement is obsolete. It does nothing.
452 Before Netpbm 10.26 (January 2005), ppmtompeg would drop trail‐
454 ing B frames from your movie, since a movie can't end with a B
455 frame. (See I Frames, P Frames, B Frames ⟨#ipb⟩ . You would
456 have to specify FORCE_ENCODE_LAST_FRAME to stop that from hap‐
457 pening and get the same function that ppmtompeg has today.
458 NIQTABLE
462 This statement specifies a custom non-intra quantization table.
463 If you don't specify this statement, ppmtompeg uses a default
464 non-intra quantization table.
465 The 8 lines immediately following NIQTABLE specify the quantiza‐
467 tion table. Each line defines a table row and consists of 8
468 integers, whitespace-delimited, which define the table columns.
469 IQTABLE
472 This is analogous to NIQTABLE, but for the intra quantization
473 table.
474 ASPECT_RATIO ratio
477 This statement specifies the aspect ratio for ppmtompeg to spec‐
478 ify in the MPEG output. I'm not sure what this is used for.
479 ratio must be 1.0, 0.6735, 0.7031, 0.7615, 0.8055, 0.8437,
481 0.8935, 0.9157, 0.9815, 1.0255, 1.0695, 1.0950, 1.1575, or
482 1.2015.
483 FRAME_RATE rate
486 This specifies the frame rate for ppmtompeg to specify in the
487 MPEG output. Some players use this value to determine the play‐
488 back rate.
489 rate must be 23.976, 24, 25, 29.97, 30, 50, 59.94, or 60.
491 BIT_RATE rate
494 This specifies the bit rate for Constant Bit Rate (CBR) encod‐
495 ing.
496 rate must be an integer.
498 BUFFER_SIZE size
501 This specifies the value ppmtompeg is to specify in the MPEG
502 output for the Video Buffering Verifier (VBV) buffer size needed
503 to decode the sequence.
504 A Video Verifying Buffer is a buffer in which a decoder keeps
506 the decoded bits in order to match the uneven speed of the
507 decoding with the required constant playback speed.
508 As ppmtompeg encodes the image, it simulates the decoding
510 process in terms of how many bits would be in the VBV as each
511 frame gets decoded, assuming a VBV of the size you indicate.
512 If you specify the WARN_VBV_UNDERFLOW statement, ppmtompeg
514 issues a warning each time the simulation underflows the buffer,
515 which suggests that an underflow would occur on playback, which
516 suggests the buffer is too small.
517 If you specify the WARN_VBV_OVERFLOW statement, ppmtompeg issues
519 a warning each time the simulation overflows the buffer, which
520 suggests that an overflow would occur on playback, which sug‐
521 gests the buffer is too small.
522 WARN_VBV_UNDERFLOW
525 WARN_VBV_OVERFLOW
527 See BUFFER_SIZE.
528 These options were new in Netpbm 10.26 (January 2005). Before
530 that, ppmtompeg issued the warnings always.
531 The following statements apply only to parallel operation:
536 PARALLEL
541 This statement, paired with END PARALLEL, is what causes ppmtom‐
542 peg to operate in parallel mode. See Parallel Operation
543 ⟨#parallel⟩ .
544 END PARALLEL
547 This goes with PARALLEL.
548 PARALLEL_TEST_FRAMES n
551 The master starts off by measuring each slave's speed. It does
552 this by giving each slave n frames to encode and noting how long
553 the slave takes to finish. These are not just test frames,
554 though -- they're real frames and the results become part of the
555 output. ppmtompeg is old and measures time in undivided sec‐
556 onds, so to get useful timings, specify enough frames that it
557 will take at least 5 seconds to process them. The default is
558 10.
559 If you specify FORCE_I_ALIGN, ppmtompeg will increase the test
561 frames value enough to maintain the alignment.
562 If there aren't enough frames for every slave to have the indi‐
564 cated number of test frames, ppmtompeg will give some slaves
565 fewer.
566 PARALLEL_TIME_CHUNKS t
570 When you specify this statement, the master attempts to feed
571 work to the slaves in chunks that take t seconds to process. It
572 uses the speed measurement it made when it started up (see PAR‐
573 ALLEL_TEST_FRAMES) to decide how many frames to put in the
574 chunk. This statement obviously doesn't affect the first batch
575 of work sent to each slave, which is the one used to measure the
576 slave's speed.
577 Smaller values of t increase communication, but improve load
579 balancing. The default is 30 seconds.
580 You may specify only one of PARALLEL_TIME_CHUNKS, PARAL‐
582 LEL_CHUNK_TAPER, and PARALLEL_PERFECT. PARALLEL_CHUNK_TAPER is
583 usually best.
584 PARALLEL_CHUNK_TAPER
587 When you specify this statement, the master distributes work
588 like with PARALLEL_TIME_CHUNKS, except that the master chooses
589 the number of seconds for the chunks. It starts with a large
590 number and, as it gets closer to finishing the job, reduces it.
591 That way, it reduces scheduling overhead when precise scheduling
592 isn't helpful, but still prevents a slave from finishing early
593 after all the work has already been handed out to the other
594 slaves, and then sitting idle while there's still work to do.
595 You may specify only one of PARALLEL_TIME_CHUNKS, PARAL‐
597 LEL_CHUNK_TAPER, and PARALLEL_PERFECT. PARALLEL_CHUNK_TAPER is
598 usually best.
599 PARALLEL_PERFECT
603 If this statement is present, ppmtompeg schedules on the assump‐
604 tion that each machine is about the same speed. The master will
605 simply divide up the frames evenly between the slaves -- each
606 slave gets the same number of frames. If some slaves are faster
607 than others, they will finish first and remain idle while the
608 slower slaves continue.
609 This has the advantage of minimal scheduling overhead. Where
611 slaves have different speeds, though, it makes inefficient use
612 of the fast ones. Where slaves are the same speed, it also has
613 the disadvantage that they all finish at the same time and feed
614 their output to the single Combine Server in a burst, which
615 makes less efficient use of the Combine Server and thus can
616 increase the total elapsed time.
617 You may specify only one of PARALLEL_TIME_CHUNKS, PARAL‐
619 LEL_CHUNK_TAPER, and PARALLEL_PERFECT. PARALLEL_CHUNK_TAPER is
620 usually best.
621 RSH remote_shell_command
624 ppmtompeg executes the shell command remote_shell_command to
625 start a process on another machine. The default command is rsh,
626 and whatever command you specify must have compatible semantics.
627 ssh is usually compatible. The command ppmtompeg uses is one
628 like this: ssh remote.host.com -l username shellcommand.
629 Be sure to set up .rhosts files or SSH key authorizations where
631 needed. Otherwise, you'll have to type in passwords.
632 On some HP machines, rsh is the restricted shell, and you want
634 to specify remsh.
635 FORCE_I_ALIGN
638 This statement forces each slave to encode a chunk of frames
639 which is a multiple of the pattern length (see PATTERN). Since
640 the first frame in any pattern is an I frame, this forces each
641 chunk encoded by a slave to begin with an I frame.
642 This document used to say there was an argument to FORCE_I_ALIGN
644 which was the number of frames ppmtompeg would use (and was
645 required to be a multiple of the pattern length). But ppmtompeg
646 has apparently always ignored that argument, and it does now.
647 KEEP_TEMP_FILES
650 This statement causes ppmtompeg not to delete the temporary
651 files it uses to transmit encoded frames to the combine server.
652 This means you will be left with a file for each frame, the same
653 as you would get with the -frames option.
654 This is mostly useful for debugging.
656 This works only if you're using a shared filesystem to communi‐
658 cate between the servers.
659 This option was new in Netpbm 10.26 (January 2005).
661 Parameter File Notes
667 If you use the -combine_gops option, then you need to specify only the
668 SIZE and OUTPUT values in the parameter file. In addition, the parame‐
669 ter file may specify input GOP files in the same manner as normal input
670 files -- except instead of using INPUT_DIR, INPUT, and END_INPUT, use
671 GOP_INPUT_DIR, GOP_INPUT, and GOP_END_INPUT. If you specify no input
672 GOP files, then ppmtompeg uses by default the output file name with
673 suffix .gop.gop_num, with gop_num starting from 0, as the input files.
674 If you use the -combine_frames option, then you need to specify only
676 the SIZE, GOP_SIZE, and OUTPUT values in the parameter file. In addi‐
677 tion, the parameter file may specify input frame files in the same man‐
678 ner as normal input files -- except instead of using INPUT_DIR, INPUT,
679 and END_INPUT, use FRAME_INPUT_DIR, FRAME_INPUT, and FRAME_END_INPUT.
680 If no input frame files are specified, then the default is to use the
681 output file name with suffix .frame.frame_num, with frame_num starting
682 from 0, as the input files.
683 Any number of spaces and tabs may come between each option and value.
685 Lines beginning with # are ignored. Any other lines are ignored except
686 for those between INPUT and END_INPUT. This allows you to use the same
687 parameter file for normal usage and for -combine_gops and -com‐
688 bine_frames.
689 The file format is case-sensitive so all keywords should be in upper
691 case.
692 The statements may appear in any order, except that the order within a
694 block statement (such as INPUT ... END INPUT) is significant.
695 ppmtompeg is prepared to handle up to 16 B frames between reference
697 frames when encoding with input from stdin. (To build a modified ppm‐
698 tompeg with a higher limit, change the constant B_FRAME_RUN in frame.c
699 and recompile).
700
703 Qscale
704 The quantization scale values (qscale) give a trade-off between quality
705 and compression. Using different Qscale values has very little effect
706 on speed. The qscale values can be set separately for I, P, and B
707 frames.
708 You select the qscale values with the IQSCALE, PQSCALE, and BSCALE
710 parameter file statements.
711 A qscale value is an integer from 1 to 31. Larger numbers give better
713 compression, but worse quality. In the following, the quality numbers
714 are peak signal-to-noise ratio, defined as: signal-to-noise formula
715 where MSE is the mean squared error.
716 Flower garden tests:
719 Qscale vs Quality
721 ────────────────────────────────────────
723 Qscale I Frames P Frames B Frames
724 1 43.2 46.3 46.5
725 6 32.6 34.6 34.3
726 11 28.6 29.5 30.0
727 16 26.3 26.8 28.6
728 21 24.7 25.0 27.9
729 26 23.5 23.9 27.5
730 31 22.6 23.0 27.3
731 Qscale vs Compression
733 ────────────────────────────────────────
735 Qscale I Frames P Frames B Frames
736 1 2 2 2
737 6 7 10 15
738 11 11 18 43
739 16 15 29 97
740 21 19 41 173
741 26 24 56 256
742 31 28 73 330
743 Search Techniques
747 There are several different motion vector search techniques available.
748 There are different techniques available for P frame search and B frame
749 search. Using different search techniques present little difference in
750 quality, but a large difference in compression and speed.
751 There are 4 types of P frame search: Exhaustive, TwoLevel, SubSample,
754 and Logarithmic.
755 There are 3 types of B frame search: Exhaustive, Cross2, and Simple.
758 The recommended search techniques are TwoLevel and Logarithmic for P
760 frame search, and Cross2 and Simple for B frame search. Here are some
761 numbers comparing the different search methods:
762 P frame Motion Vector Search (Normalized)
764 ─────────────────────────────────────────────────────────────────────
766 Technique Compression 1 Speed 2 Quality 3
767 ⟨#smallbetter⟩ ⟨#largefaster⟩ ⟨#largebetter⟩
768 Exhaustive 1000 1000 1000
769 SubSample 1008 2456 1000
770 TwoLevel 1009 3237 1000
771 Logarithmic 1085 8229 998
773 B frame Motion Vector Search (Normalized)
775 ────────────────────────────────────────────────────────────────────
777 Technique Compression 1 Speed 2 Quality 3
778 ⟨#smallbetter⟩ ⟨#largefaster⟩ ⟨#largebetter⟩
779 Exhaustive 1000 1000 1000
780 Cross2 975 1000 996
781 Simple 938 1765 991
782 1Smaller numbers are better compression.
784 2Larger numbers mean faster execution.
786 3Larger numbers mean better quality.
788 For some reason, Simple seems to give better compression, but it
790 depends on the image sequence.
791 Select the search techniques with the PSEARCH_ALG and BSEARCH_ALG
793 parameter file statements.
794 Group Of Pictures (GOP)
798 A Group of Pictures (GOP) is a roughly independently decodable sequence
799 of frames. An MPEG video stream is made of one or more GOP's. You may
800 specify how many frames should be in each GOP with the GOP_SIZE parame‐
801 ter file statement. A GOP always starts with an I frame.
802 Instead of encoding an entire sequence, you can encode a single GOP.
804 To do this, use the -gop command option. You can later join the
805 resulting GOP files at any time by running ppmtompeg with the -com‐
806 bine_gops command option.
807 Slices
811 A slice is an independently decodable unit in a frame. It can be as
812 small as one macroblock, or it can be as big as the entire frame. Bar‐
813 ring transmission error, adding slices does not change quality or
814 speed; the only effect is slightly worse compression. More slices are
815 used for noisy transmission so that errors are more recoverable. Since
816 usually errors are not such a problem, we usually just use one slice
817 per frame.
818 Control the slice size with the SLICES_PER_FRAME parameter file state‐
821 ment.
822 Some MPEG playback systems require that each slice consist of whole
824 rows of macroblocks. If you are encoding for this kind of player, if
825 the height of the image is H pixels, then you should set the
826 SLICES_PER_FRAME to some number which divides H/16. For example, if
827 the image is 240 pixels (15 macroblocks) high, then you should use only
828 15, 5, 3, or 1 slices per frame.
829 Note: these MPEG playback systems are really wrong, since the MPEG
832 standard says this doesn't have to be so.
833 Search Window
838 The search window is the window in which ppmtompeg searches for motion
839 vectors. The window is a square. You can specify the size of the
840 square, and whether to allow half-pixel motion vectors or not, with the
841 RANGE and PIXEL parameter file statements.
842 I Frames, P Frames, B Frames
845 In MPEG-1, a movie is represented as a sequence of MPEG frames, each of
846 which is an I Frame, a P Frame, or a B Frame. Each represents an
847 actual frame of the movie (don't get confused by the dual use of the
848 word "frame." A movie frame is a graphical image. An MPEG frame is a
849 set of data that describes a movie frame).
850 An I frame ("intra" frame) describes a movie frame in isolation --
852 without respect to any other frame in the movie. A P frame ("predic‐
853 tive" frame) describes a movie frame by describing how it differs from
854 the movie frame described by the latest preceding I or P frame. A B
855 frame ("bidirectional" frame) describes a movie frame by describing how
856 it differs from the the movie frames described by the nearest I or P
857 frame before and after it.
858 Note that the first frame of a movie must be described by an I frame
860 (because there is no previous movie frame) and the last movie frame
861 must be described by an I or P frame (because there is no subsequent
862 movie frame).
863 Beyond that, you can choose which frames are represented by which
865 types. You specify a pattern, such as IBPBP and ppmtompeg simply
866 repeats it over and over throughout the movie. The pattern affects
867 speed, quality, and stream size. Here is a chart which shows some of
868 the trade-offs:
869 Comparison of I/P/B Frames (Normalized)
871 ────────────────────────────────────
873 Frame Type Size Speed Quality
874 I frames 1000 1000 1000
875 P frames 409 609 969
876 B frames 72 260 919
877 (this is with constant qscale)
879 A standard sequence is IBBPBBPBBPBBPBB.
882 Select the sequence with the PATTERN parameter file statement.
885 Since the last MPEG frame cannot be a B frame (see above), if the pat‐
887 tern you specify indicates a B frame for the last movie frame of the
888 movie, ppmtompeg makes it an I frame instead.
889 Before Netpbm 10.26 (January 2005), ppmtompeg instead drops the trail‐
891 ing B frames by default, and you need the FORCE_ENCODE_LAST_FRAME
892 parameter file statement to make it do this.
893 The MPEG frames don't appear in the MPEG-1 stream in the same order
895 that the corresponding movie frames appear in the movie -- the B frames
896 come after the I and P frames on which they are based. For example, if
897 the movie is 4 frames that you will represent with the pattern IBBP,
898 the MPEG-1 stream will start with an I frame describing movie frame 0.
899 The next frame in the MPEG-1 stream is a P frame describing movie frame
900 3. The last two frames in the MPEG-1 stream are B frames describing
901 movie frames 1 and 2, respectively.
902 Specifying Input and Output Files
906 Specify the input frame images with the INPUT_DIR, INPUT, END_INPUT,
907 BASE_FILE_FORMAT, SIZE, YUV_FORMAT and INPUT_CONVERT parameter file
908 statements.
909 Specify the output file with the OUTPUT parameter file statement.
911 Statistics
915 ppmtompeg can generate a variety of statistics about the encoding. See
916 the -stat, -snr, -mv_histogram, -quiet, -no_frame_summary, and
917 -bit_rate_info options.
918
922 You can run ppmtompeg on multiple machines at once, encoding the same
923 MPEG stream. When you do, the machines are used as shown in the fol‐
924 lowing diagram. We call this 'parallel mode.'
925 ppmtompeg-par.gif
927 To do parallel processing, put the statement
929 PARALLEL
931 in the parameter file, followed by a listing of the machines, one
933 machine per line, then
934 END_PARALLEL
936 Each of the machine lines must be in one of two forms. If the machine
938 has filesystem access to the input files, then the line is:
939 machine user executable
941 The executable is normally ppmtompeg (you may need to give the complete
943 path if you've built for different architectures). If the machine does
944 not have filesystem access to the input files, the line is:
945 REMOTE machine user executable parameter file
947 The -max_machines command option limits the number of machines ppmtom‐
949 peg will use. If you specify more machines in the parameter file than
950 -max_machines allows, ppmtompeg uses only the machines listed first.
951 This is handy if you want to experiment with different amounts of par‐
952 allelism.
953 In general, you should use full path file names when describing exe‐
955 cutables and parameter files. This includes the parameter file argu‐
956 ment on the original invocation of ppmtompeg.
957 All file names must be the same on all systems (so if e.g. you're using
959 an NFS filesystem, you must make sure it is mounted at the same mount‐
960 point on all systems).
961 Because not all of the processes involved in parallel operation have
963 easy access to the input files, you must specify the SIZE parameter
964 file statement when you do parallel operation.
965 The machine on which you originally invoke ppmtompeg is the master
967 machine. It hosts a 'combine server,', a 'decode server,' and a number
968 of 'i/o servers,' all as separate processes. The other machines in the
969 network (listed in the parameter file) are slave machines. Each hosts
970 a single process that continuously requests work from the master and
971 does it. The slave process does the computation to encode MPEG frames.
972 It processes frames in batches identified by the master.
973 The master uses a remote shell command to start a process on a slave
975 machine. By default, it uses an rsh shell command to do this. But use
976 the RSH parameter file statement to control this. The shell command
977 the master executes remotely is ppmtompeg, but with options to indicate
978 that it is to perform slave functions.
979 The various machines talk to each other over TCP connections. Each
981 machine finds and binds to a free TCP port number and tells its part‐
982 ners the port number. These port numbers are at least 2048.
983 Use the PARALLEL_TEST_FRAMES, PARALLEL_TIME_CHUNKS, and PARALLEL_PER‐
985 FECT parameter file statements to control the way the master divides up
986 work among the slaves.
987 Use the -nice command option to cause all slave processes to run
989 "nicely," i.e. as low priority processes. That way, this substantial
990 and long-running CPU load will have minimal impact on other, possibly
991 interactive, users of the systems.
992
995 Here is a look at ppmtompeg speed, in single-node (not parallel) opera‐
996 tion:
997 Compression Speed
999 ───────────────────────────────────────
1002 Machine Type Macroblocks per second1
1003 HP 9000/755 280
1004 DEC 3000/400 247
1005 HP 9000/750 191
1006 Sparc 10 104
1007 DEC 5000 68
1008 1A macroblock is a 16x16 pixel square
1009 The measurements in the table are with inputs and outputs via a conven‐
1011 tional locally attached filesystem. If you are using a network
1012 filesystem over a single 10 MB/s Ethernet, that constrains your speed
1013 more than your CPU speed. In that case, don't expect to get better
1014 than 4 or 5 frames per second no matter how fast your CPUs are.
1015 Network speed is even more of a bottleneck when the slaves do not have
1017 filesystem access to the input files -- i.e. you declare them REMOTE.
1018 Where I/O is the bottleneck, size of the input frames can make a big
1020 difference. So YUV input is better than PPM, and JPEG is better than
1021 both.
1022 When you're first trying to get parallel mode working, be sure to use
1024 the -debug_machines option so you can see what's going on. Also,
1025 -debug_sockets can help you diagnose communication problems.
1026
1030 · Kevin Gong - University of California, Berkeley, kev‐
1031 ing@cs.berkeley.edu
1032 · Ketan Patel - University of California, Berkeley, kpa‐
1035 tel@cs.berkeley.edu
1036 · Dan Wallach - University of California, Berkeley, dwal‐
1039 lach@cs.berkeley.edu
1040 · Darryl Brown - University of California, Berkeley, dar‐
1043 ryl@cs.berkeley.edu
1044 · Eugene Hung - University of California, Berkeley,
1047 eyhung@cs.berkeley.edu
1048 · Steve Smoot - University of California, Berkeley,
1051 smoot@cs.berkeley.edu
1052netpbm documentation 23 July 2006 Ppmtompeg User Manual(0)