| 1 |
xvid vfw help |
| 2 |
20/jan/02 |
| 3 |
|
| 4 |
*** rate control *** |
| 5 |
|
| 6 |
xvid offers a number of different rate control modes |
| 7 |
|
| 8 |
- 1pass-cbr: constant-bitrate |
| 9 |
use the bitrate slider to select the desired bitrate; refer to 'cbr constants' for advanced options. note: we actually use an average-bitrate algorithm, so the instaneous bitrate will fluxuate. |
| 10 |
|
| 11 |
- 1pass-quality |
| 12 |
use the quality slider to select a desired quality percentile. the quality% is then used to calculate an average-quantizer. |
| 13 |
|
| 14 |
- 1pass-quantizer |
| 15 |
force the codec to use a fixed quantizer for all frames. valid range: [1,31]. |
| 16 |
|
| 17 |
- 2pass |
| 18 |
a detailed explaination of 2pass can be found here: http://www.research.ibm.com/journal/rd/434/westerink.html |
| 19 |
our implementation of two pass encoding requires three steps |
| 20 |
1) 1stpass: collect compression statisics to a stats file. |
| 21 |
2) process the stats using gordian knot, xihprom, etc. |
| 22 |
3) 2ndpass: compress the video using both the original 1stpass stats file, and the normalized stats file. |
| 23 |
|
| 24 |
|
| 25 |
*** encoder tuning *** |
| 26 |
|
| 27 |
- max keyframe interval |
| 28 |
the maximum interval between keyframes. |
| 29 |
|
| 30 |
- motion search precision |
| 31 |
controls the speed/quality of the motion search algorithm. range [0,6]. |
| 32 |
lower values are faster but reduce compressibility. |
| 33 |
higher values are slower but increase compressibility. |
| 34 |
|
| 35 |
- quantization type |
| 36 |
quantization is the process of removing detail. there are two types: |
| 37 |
h.263: used by divx4.x and h263 encoders. |
| 38 |
mpeg: used by the microsoft mpeg-4 encoder, mpeg-1/2. |
| 39 |
|
| 40 |
mpeg quantization gives sharper looking images, but can introduce grain/noise. while h.263 gives softer looking images, removes noise (but sometimes too much) and is slightly faster. |
| 41 |
|
| 42 |
|
| 43 |
*** cbr constants *** |
| 44 |
|
| 45 |
these apply only to 1pass-cbr mode. |
| 46 |
|
| 47 |
- average peroid |
| 48 |
the number of frames over which the average bitrate is maintained. |
| 49 |
|
| 50 |
- reaction peroid |
| 51 |
(?) the number of frames inwhich we react to sudden change in bitrate. |
| 52 |
|
| 53 |
- up/down ratio |
| 54 |
(?) a percentile representing how fast/much we react. |
| 55 |
|
| 56 |
|
| 57 |
*** debug *** |
| 58 |
|
| 59 |
- min/max quantizer. |
| 60 |
lets you limit the flexibility of the encoder. |
| 61 |
a minquant=1 can sometimes cause problems with 1pass-cbr and 2pass. |
| 62 |
|
| 63 |
- luminance masking |
| 64 |
enabled per-macroblock quantizer decision. very experimental |
| 65 |
|
| 66 |
- discard encoded bitstream on 1st-pass |
| 67 |
by default xvid discards the compressed bitstream during the the 1st pass of a two pass encoding session. example: in windows you end up with an empty .avi file. if this box is unchecked the bitstream is kept. |
| 68 |
note: the 1st compressed bitstream can large. |
| 69 |
|
| 70 |
this can be used to perform a 'fast 2pass' encoding. |
| 71 |
1) perform 1st pass encoding of your source material with discard unchecked. |
| 72 |
2) process the stats file |
| 73 |
3) open the bitstream created in the first pass, and perform the 2nd pass on it. |
| 74 |
note: there is some quality loss. judge it for yourself. |
| 75 |
|
| 76 |
- perform dummy 2nd-pass |
| 77 |
this can be used to check the accuracy of your stats file processor/calculator. it instructs xvid to create a dummy/junk bitstream exactly to the size described in the processed stats file. |
| 78 |
|
| 79 |
*** cpu *** |
| 80 |
|
| 81 |
provides auto/manual control over cpu-feature optimizations. |
| 82 |
|
| 83 |
|
| 84 |
*** about *** |
| 85 |
|
| 86 |
build timestamp, etc. |
