1
0
mirror of https://github.com/mpv-player/mpv synced 2024-12-19 21:31:52 +00:00
mpv/DOCS/tech/vop.txt
diego 2f095b4652 Spellchecked, slightly reworded, parts corrected.
git-svn-id: svn://svn.mplayerhq.hu/mplayer/trunk@9680 b3059339-0415-0410-9bf9-f77b7e298cf2
2003-03-24 17:40:44 +00:00

242 lines
10 KiB
Plaintext

Video Filters
=============
Video filters are plugin-like code modules implementing the interface
defined in vf.h.
Basically it means video output manipulation, i.e. these plugins can
modify the image and the image properties (size, colorspace etc) between
the video decoders (vd.h) and the output layer (libvo or video encoders).
The actual API is a mixture of the video decoder (vd.h) and libvo
(video_out.h) APIs.
main differences:
- vf plugins may be "loaded" multiple times, with different parameters
and context - it's new in MPlayer, old APIs weren't reentrant.
- vf plugins don't have to implement all functions - all functions have a
'fallback' version, so the plugins only override these if wanted.
- Each vf plugin has its own get_image context, and they can interchange
images/buffers using these get_image/put_image calls.
API details:
-- TO DO -- (see vf.h for API and vf_*.c for examples)
Current plugins:
================
-vf crop[=w:h:x:y]
Simple cropping plugin.
w,h (cropped width,height) defaults to original width,height
x,y (position of cropped subimage on the original image) defaults to center
MPI: EXPORT only, using stride manipulation
-vf expand[=w:h:x:y:o]
Expanding _and_ (optional) OSD rendering plugin.
w,h (expanded width,height) defaults (-1) to original width,height
x,y (position of original image on the expanded image) defaults (-1) to center
o (0=disable/1=enable SUB/OSD rendering) defaults (0) to disabled
MPI: DR (if possible) or copy
Special: mpcodecs core uses it to solve stride restrictions between filters
MEncoder uses (autoload) it to render SUB/OSD
-vf flip
Flips image upside-down (vertical mirroring)
No parameters.
MPI: DR (if possible) or EXPORT, using stride manipulation
Special: dec_video autoloads it when flipping is required and libvo can't do
-vf mirror
Horizontal mirroring
No parameters.
MPI: TEMP
-vf rectangle[=w:h:x:y]
Draw a rectangle. Useful for testing crop plugin parameters.
w,h (rectangle's width and height) defaults (-1) to maximum
possible width while keeping the boundaries visible.
x,y (rectangle's top left corner position) defaults (-1) to upper
left most position.
The plugin responds to the input.conf directive "change_rectangle"
that takes two parameters. The first parameter can be 0 for w, 1
for h, 2 for x or 3 for y. The second parameter is the amount to
change the designated rectangle boundary.
MPI: TEMP, accepts stride
-vf rotate[=x]
Rotate image +/- 90 degrees
Optional 'x' parameter (0..3) controls horizontal and vertical mirroring
MPI: TEMP
-vf scale[=w:h[:c[:p]]]
Software scaling (zoom) _and_ yuv<->rgb colorspace conversion
w,h (new width/height after scaling) defaults to original width,height
Note: If -zoom is used, and underlying filters (including libvo) are
incapable of scaling, then it defaults to d_width/d_height!
note 2: w/h values -1 means original width/height, 0 means scaled
d_width/d_height. -2/-3 means calculating w or h using the other
dimension and the original (-3) or prescaled (-2) aspect ratio.
c chroma skipping
0 -> use all available input lines for chroma
1 -> use only every 2. input line for chroma
2 -> use only every 4. input line for chroma
3 -> use only every 8. input line for chroma
p scaling parameter (depends upon the scaling method used)
for -sws 2 (bicubic) it is sharpness (0 (soft) - 100 (sharp))
for -sws 7 (gaussian) it is sharpness (0 (soft) - 100 (sharp))
for -sws 9 (lanczos) it is filter length (1 - 10)
MPI: TEMP, accepts stride
Special: dec_video and mpcodecs core autoloads it for colorspace conv.
-vf yuy2
Forced software YV12/I420 -> YUY2 conversion
(useful for video cards/drivers with slow YV12 but fast YUY2 support)
MPI: TEMP, accepts stride
-vf rgb2bgr[=swap]
RGB 24/32 <-> BGR 24/32 colorspace conversion (default) or
RGB 24/32 <-> RGB 24/32 conversion with R<->B swapping ('swap' option)
MPI: TEMP, accepts stride
-vf palette
RGB/BGR 8 -> RGB/BGR 15/16/24/32 colorspace conversion using palette
MPI: TEMP, accepts stride
-vf format[=fourcc]
_restrict_ the list of supported colorspaces (query-format()) to a single,
given fourcc. The 'fourcc' option defaults to 'yuy2', but may be any
format name, like rgb15, bgr24, yv12 etc...
Note again that it does NOT do any conversion, it just limits the format
list for the _next_ plugin.
MPI: passthru
-vf pp=[<filterName>[:<option>[:<option>...]][/[-]<filterName>[:<option>...]]...]
postprocessing filter. (useful for codecs without built-in
postprocessing, like libmpeg12 or libavcodec)
see `mplayer -vf pp=help`
MPI: DR (if possible) or TEMP
Special: dec_video autoloads it if -pp option is used but codec can't do pp.
-vf lavc[=quality:fps]
Fast software YV12->MPEG1 conversion, useful for DXR3/DVB drivers
It uses libavcodec. Currently faster and better quality than -vf fame!
Param quality: fixed qscale (1<=quality<32) or bitrate (32<=quality KBits)
Param fps: force output fps (float value). If not give, or 0, it's
autodetected based on height (240,480->29.97fps, others -> 25fps)
MPI: EXPORT special
-vf fame
Fast software YV12->MPEG1 conversion, useful for DXR3/DVB drivers
It uses libfame.
MPI: EXPORT special
-vf dvbscale[=aspect]
Setup scaling to the optimal values for the DVB card.
(use HW for X-scaling, calc sw Y scaling to keep aspect)
The 'aspect' parameter controls aspect ratio, it should be calculated
as aspect=DVB_HEIGHT*ASPECTRATIO, default is 576*(4/3)=768
(for 16:9 TV set it to 576*(16/9)=1024)
It's only useful together with expand+scale:
-vf dvbscale,scale=-1:0,expand=-1:576:-1:-1:1,lavc
MPI: passthru
-vf cropdetect[=limit]
It's a special filter, it doesn't alter the image itself, but detects
(using some kind of heuristics) the black borders/bands of the image,
and prints the optimal -vf crop= parameters to the stdout/console.
It calculates the average value of pixel luminances for a line, and
depending on the result it's either black band (result<=limit) or
picture (result>limit). Optional parameter 'limit' defaults to 24.
MPI: EXPORT
-vf test[=first frame number]
generate various test patterns
MPI: TEMP, accepts stride
-vf noise[=lumaNoise[u][t|a][h]:chromaNoise[u][t|a][h]
add noise
0<= lumaNoise, chromaNoise <=100
u uniform noise (gaussian otherwise)
t temporal noise (noise pattern changes between frames)
a averaged temporal (smoother, but a lot slower)
h high quality (slightly better looking, slightly slower)
p mix random noise with a (semi)regular pattern
MPI: DR (if possible) or TEMP, accepts stride
-vf eq[=brightness:contrast]
software equalizer, for use with cards that don't support
brightness and contrast controls in hardware. It might also be
useful with MEncoder, either for fixing poorly captured movies, or
for slightly reducing contrast to mask artifacts and get by with
lower bitrates. Initial values in the range -100..100 may be
given on the command line, and the eq filter is controllable
interactively just like with normal hardware equalizer controls.
-vf eq2[=gamma:contrast:brightness:saturation:r_gamma:g_gamma:b_gamma]
advanced software equalizer, for use with cards that don't support
brightness/contrast/saturation and channel gamma controls in hardware.
Initial values in the range -2.0 .. 2.0 may be given on the command line,
and the eq2 filter is controllable interactively just like with normal
hardware equalizer controls or -vf eq. Defaults: 1:1:0:1:1:1:1
This filter uses MMX only if available and all gamma == 1.0, otherwise
LUT (look-up table) is being used.
-vf unsharp=l|cWxH:amount[:l|cWxH:amount]
unsharp mask / gaussian blur.
l apply effect on luma components
c apply effect on chroma components
WxH width and height of the matrix, odd sized in both directions
min = 3x3, max = 13x11 or 11x13
usually you will use something between 3x3 and 7x7
amount relative amount of sharpness / blur to add to the image
amount < 0 = blur, amount > 0 = sharpen
usually you will use something between -1.5 and 1.5
MPI: DR (if possible) or TEMP, accepts stride
-vf swapuv
swap U & V plane
MPI: EXPORT
-vf il=[d|i][s][:[d|i][s]]
(de)interleaves lines
d deinterleave
i interleave
s swap fields (exchange even & odd lines)
The goal of this filter is to add the ability to process interlaced images
pre-field without deinterlacing them. You can filter your interlaced DVD
and play it on a TV without breaking the interlacing.
While deinterlacing (with the postprocessing filter) removes interlacing
permanently (by smoothing, averaging, etc) deinterleaving splits the
frame into 2 fields (so called half pictures), so you can process (filter)
them independently and then re-interleave them.
MPI: TEMP, accepts stride
-vf boxblur=radius:power[:radius:power]
radius size of the filter
power how often the filter should be applied
MPI: TEMP, accepts stride
-vf sab=radius:prefilter:colorDiff[:radius:prefilter:colorDiff]
radius blur filter strength (~0.1-4.0) (slower if larger)
prefilter prefilter strength (~0.1-2.0)
colorDiff how different the pixels are allowed to be considered (~0.1-100.0)
-vf smartblur=radius:stregth:threshold[:radius:stregth:threshold]
radius blur filter strength (~0.1-5.0) (slower if larger)
strength (0.0-1.0) -> blur, (-1.0-0.0) -> sharpen
threshold 0 -> filter all, (0-30) -> filter flat areas, (-30-0) -> filter edges
-vf perspective=x0:y0:x1:y1:x2:y2:x3:y3:t
x0,y0,... coordinates of the top left, top right, bottom left, bottom right corners
t 0-> linear, 1->cubic resampling
-vf denoise3d=[luma:chroma:time]
luma spatial luma strength (default = 4)
chroma spatial chroma strength (default = 3)
time temporal strength (default = 6)
This filter aims to reduce image noise producing smooth images and
making still images really still. (it should enhance compressibility)
Can be given from 0 to 3 parameters. Parameters not given will be
inferred from the given ones. (i.e. providing only one of them acts as a
"scale factor")