Re: [FFmpeg-devel] [PATCH] avcodec/mjpegdec: support weird RGB subsampling with progressive

[Leo Izen <leo.izen@gmail.com>]

On 3/15/23 23:11, Andreas Rheinhardt wrote:
> Leo Izen:
>> This allows weird subsampling with progressive JPEGs to be decoded,
>> such as full-RG and only B subsampled.
>> ---
>>   libavcodec/mjpegdec.c | 38 ++++++++++++++++++++++++++++----------
>>   1 file changed, 28 insertions(+), 10 deletions(-)
>>
>> diff --git a/libavcodec/mjpegdec.c b/libavcodec/mjpegdec.c
>> index c833d66c4d..062730f3e1 100644
>> --- a/libavcodec/mjpegdec.c
>> +++ b/libavcodec/mjpegdec.c
>> @@ -571,10 +571,15 @@ int ff_mjpeg_decode_sof(MJpegDecodeContext *s)
>>           case 0x22221100:
>>           case 0x22112200:
>>           case 0x11222200:
>> -            if (s->bits <= 8) s->avctx->pix_fmt = s->cs_itu601 ? AV_PIX_FMT_YUV444P : AV_PIX_FMT_YUVJ444P;
>> -            else
>> +            if (s->bits > 8)
>>                   goto unk_pixfmt;
>> -            s->avctx->color_range = s->cs_itu601 ? AVCOL_RANGE_MPEG : AVCOL_RANGE_JPEG;
>> +            if (s->adobe_transform == 0 || s->component_id[0] == 'R' &&
>> +                    s->component_id[1] == 'G' && s->component_id[2] == 'B') {
>> +                s->avctx->pix_fmt = AV_PIX_FMT_GBRP;
>> +            } else {
>> +                s->avctx->pix_fmt = s->cs_itu601 ? AV_PIX_FMT_YUV444P : AV_PIX_FMT_YUVJ444P;
>> +                s->avctx->color_range = s->cs_itu601 ? AVCOL_RANGE_MPEG : AVCOL_RANGE_JPEG;
>> +            }
>>               break;
>>           case 0x11000000:
>>           case 0x13000000:
>> @@ -673,10 +678,6 @@ int ff_mjpeg_decode_sof(MJpegDecodeContext *s)
>>               avpriv_report_missing_feature(s->avctx, "Lowres for weird subsampling");
>>               return AVERROR_PATCHWELCOME;
>>           }
>> -        if ((AV_RB32(s->upscale_h) || AV_RB32(s->upscale_v)) && s->progressive && s->avctx->pix_fmt == AV_PIX_FMT_GBRP) {
>> -            avpriv_report_missing_feature(s->avctx, "progressive for weird subsampling");
>> -            return AVERROR_PATCHWELCOME;
>> -        }
>>           if (s->ls) {
>>               memset(s->upscale_h, 0, sizeof(s->upscale_h));
>>               memset(s->upscale_v, 0, sizeof(s->upscale_v));
>> @@ -1697,9 +1698,6 @@ int ff_mjpeg_decode_sos(MJpegDecodeContext *s, const uint8_t *mb_bitmask,
>>           s->h_scount[i]  = s->h_count[index];
>>           s->v_scount[i]  = s->v_count[index];
>>   
>> -        if((nb_components == 1 || nb_components == 3) && s->nb_components == 3 && s->avctx->pix_fmt == AV_PIX_FMT_GBR24P)
>> -            index = (index+2)%3;
>> -
>>           s->comp_index[i] = index;
>>   
>>           s->dc_index[i] = get_bits(&s->gb, 4);
>> @@ -2745,6 +2743,26 @@ the_end:
>>               }
>>           }
>>       }
>> +
>> +    if (s->avctx->pix_fmt == AV_PIX_FMT_GBRP) {
>> +        int w = s->picture_ptr->width;
>> +        int h = s->picture_ptr->height;
>> +        av_assert0(s->nb_components == 3);
>> +        for (int i = 0; i < h; i++) {
>> +            uint8_t *dst[3];
>> +            for (int index = 0; index < 3; index++)
>> +                dst[index] = s->picture_ptr->data[index] + i * s->picture_ptr->linesize[index];
>> +            for (int j = 0; j < w; j++) {
>> +                uint8_t r = dst[0][j];
>> +                uint8_t g = dst[1][j];
>> +                uint8_t b = dst[2][j];
>> +                dst[0][j] = g;
>> +                dst[1][j] = b;
>> +                dst[2][j] = r;
>> +            }
>> +        }
>> +    }
> 
> Is there a reason you are not just swapping the pointers?

No particular reason, other than I tried that and messed it up when 
testing (probably because I didn't do it properly). I'll send a v2 and 
swap these.

- Leo Izen (thebombzen)

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

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