Jan Ekström (12020-09-16):
> This value - while it looks like the actual range of the content -
> is nothing but the internal value of swscale.
> 
> Thus, if we have RGB content, force the value to 1. Swscale will
> ignore it, but at least the value of the output AVFrame will now
> properly be "full range" instead of "limited range", as it is right
> now.
> ---
>  libavfilter/vf_scale.c | 17 +++++++++++++++++
>  1 file changed, 17 insertions(+)
> 
> diff --git a/libavfilter/vf_scale.c b/libavfilter/vf_scale.c
> index 58eee96744..12df27edf4 100644
> --- a/libavfilter/vf_scale.c
> +++ b/libavfilter/vf_scale.c
> @@ -751,6 +751,15 @@ scale:
>          || scale->out_range != AVCOL_RANGE_UNSPECIFIED) {
>          int in_full, out_full, brightness, contrast, saturation;
>          const int *inv_table, *table;

> +        const AVPixFmtDescriptor *out_desc = 
> av_pix_fmt_desc_get(out->format);
> +        if (!out_desc) {
> +            av_log(ctx, AV_LOG_ERROR,
> +                   "Failed to get the pixel format descriptor for format 
> %d!\n",
> +                   out->format);
> +            av_frame_free(&in);
> +            av_frame_free(frame_out);
> +            return AVERROR_INVALIDDATA;
> +        }

Since the format is negotiated, I think it can be considered a severe
internal bug that out_desc is not found, i.e. checked by an assert.

>  
>          sws_getColorspaceDetails(scale->sws, (int **)&inv_table, &in_full,
>                                   (int **)&table, &out_full,
> @@ -768,7 +777,15 @@ scale:
>          else if (in_range != AVCOL_RANGE_UNSPECIFIED)
>              in_full  = (in_range == AVCOL_RANGE_JPEG);
>          if (scale->out_range != AVCOL_RANGE_UNSPECIFIED)
> +            // note: this can be silently overridden by
> +            //       sws_setColorspaceDetails for non-YCbCr formats
>              out_full = (scale->out_range == AVCOL_RANGE_JPEG);
> +        else if (out_desc->flags & AV_PIX_FMT_FLAG_RGB)
> +            // the range values received from swscale are its internal
> +            // identifiers, and will not be nonzero for any sort of RGB.
> +            // thus, if we do not set it to nonzero ourselves, this will
> +            // be incorrect.
> +            out_full = 1;
>  
>          sws_setColorspaceDetails(scale->sws, inv_table, in_full,
>                                   table, out_full,

With you explanation, I think it is ok, but I do not know the topic
enough to be sure.

Regards,

-- 
  Nicolas George

Attachment: signature.asc
Description: PGP signature

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".

Reply via email to