#4805 closed defect (fixed)
swscale does not process pixels when only color-matrix differs between input and output
| Reported by: | rmk | Owned by: | |
|---|---|---|---|
| Priority: | normal | Component: | swscale |
| Version: | git-master | Keywords: | |
| Cc: | mirh | Blocked By: | |
| Blocking: | Reproduced by developer: | yes | |
| Analyzed by developer: | no |
Description
swscale does not seem to process pixels when only color-matrix differs between input and output (i.e. pixel format and size remain the same).
It would be great to be able to use swscale also just for colorspace transformations like for bt601<->bt709 conversions.
Observed this as api user but verified it by converting two files as follows and comparing them via framemd5:
ffmpeg -i fcp7_xdcam_ex_1080_25p_1s.mov -vf scale=in_color_matrix=bt709:out_color_matrix=bt601 -c:v ffv1 -an scale_test_bt601.nut
ffmpeg -i xdcam_ex/fcp7_xdcam_ex_1080_25p_1s.mov -c:v ffv1 -an scale_test_unscaled.nut
The frames of the resulting files are identical.
Change History (9)
comment:1 by , 9 years ago
follow-up: 4 comment:2 by , 9 years ago
We have the colormatrix filter for this kind of conversions.
-vf colormatrix=src=bt709:dst=bt601
comment:3 by , 9 years ago
Hard to tell what's going on here without knowing the parameters set on libswscale, but I'd say this definitely should work. Without special filter.
It's quite possible that libswscale doesn't implement this at all, though.
comment:4 by , 9 years ago
Replying to heleppkes:
We have the colormatrix filter for this kind of conversions.
-vf colormatrix=src=bt709:dst=bt601
I know but
- it's GPL (in my case that means I cannot use it),
- it's very limited as far as pixel format is concerned, compared to swscale
- it's more limited as far as color space support is concerned, compared to swscale
- it seems like a good idea to only do the maintenance for coefficients etc. in one place and since the swscale api contract (as well as the vf_scale options) seems well-suited for the task, it is probably surprising for users to find that the functionality is not in that place (although there might be reasons as far as code complexity is concerned, that I don't know, not really knowing the impl code)
I can only guess that the reason the colormatrix filter is there, was that someone else ran into this limitation, maybe knew the colormatrix support from ffmbc and decided to rather port than tackle the more complex task of making swscale support it (which is completely legitimate, of course).
follow-up: 6 comment:5 by , 9 years ago
| Component: | undetermined → swscale |
|---|---|
| Reproduced by developer: | set |
| Status: | new → open |
| Type: | enhancement → defect |
| Version: | unspecified → git-master |
See also http://thread.gmane.org/gmane.comp.video.ffmpeg.user/57005/focus=57093 and my incomplete patch in http://ffmpeg.org/pipermail/ffmpeg-devel/2015-May/173223.html - the second hunk doesn't work for some reason, the first hunk now looks like this and should be necessary afaict:
diff --git a/libavfilter/vf_scale.c b/libavfilter/vf_scale.c
index 9698bfb..be061e2 100644
--- a/libavfilter/vf_scale.c
+++ b/libavfilter/vf_scale.c
@@ -350,6 +350,7 @@ static int config_props(AVFilterLink *outlink)
scale->isws[0] = scale->isws[1] = scale->sws = NULL;
if (inlink0->w == outlink->w &&
inlink0->h == outlink->h &&
+ scale->in_range == scale->out_range &&
inlink0->format == outlink->format)
;
else {
comment:6 by , 9 years ago
Replying to cehoyos:
See also http://thread.gmane.org/gmane.comp.video.ffmpeg.user/57005/focus=57093
was this ever made into a ticket? because I think I just also ran into that.
comment:8 by , 9 years ago
| Resolution: | → fixed |
|---|---|
| Status: | open → closed |
This was fixed by Michael in 8e05f9217ae567a26ea71dd29acd79b1631ae593 (and possibly a subsequent commit).
Thank you for the report, sorry about mixing up range and matrix here.
comment:9 by , 3 years ago
| Cc: | added |
|---|
FWIW this only fixed the "manually specified" case though #9693
Full command lines and output:
Roberts-MacBook-Pro:colortest krueger$ ffmpeg -i ~/lesspain/samples/software/fcp_7/xdcam_ex/fcp7_xdcam_ex_1080_25p_1s.mov -vf scale=in_color_matrix=bt709:out_color_matrix=bt601 -c:v ffv1 -an -y scale_test_bt601.nut
ffmpeg version N-74034-gce46627 Copyright (c) 2000-2015 the FFmpeg developers
[mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fb62380da00] nclc: pri 1 trc 1 matrix 1Guessed Channel Layout for Input Stream #0.1 : stereo
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/Users/krueger/lesspain/samples/software/fcp_7/xdcam_ex/fcp7_xdcam_ex_1080_25p_1s.mov':
Output #0, nut, to 'scale_test_bt601.nut':
Stream mapping:
Press [q] to stop, ? for help
frame= 25 fps=0.0 q=-0.0 Lsize= 3715kB time=00:00:01.00 bitrate=30435.4kbits/s
video:3714kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.034393%
Roberts-MacBook-Pro:colortest krueger$ ffmpeg -i ~/lesspain/samples/software/fcp_7/xdcam_ex/fcp7_xdcam_ex_1080_25p_1s.mov -c:v ffv1 -an -y scale_test_unscaled.nut
ffmpeg version N-74034-gce46627 Copyright (c) 2000-2015 the FFmpeg developers
[mov,mp4,m4a,3gp,3g2,mj2 @ 0x7fa1bc00da00] nclc: pri 1 trc 1 matrix 1Guessed Channel Layout for Input Stream #0.1 : stereo
Input #0, mov,mp4,m4a,3gp,3g2,mj2, from '/Users/krueger/lesspain/samples/software/fcp_7/xdcam_ex/fcp7_xdcam_ex_1080_25p_1s.mov':
Output #0, nut, to 'scale_test_unscaled.nut':
Stream mapping:
Press [q] to stop, ? for help
frame= 25 fps=0.0 q=-0.0 Lsize= 3715kB time=00:00:01.00 bitrate=30435.4kbits/s
video:3714kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.034393%