[FFmpeg-devel] [PATCH v3] lavc/rawdec: Use AV_PIX_FMT_PAL8 for 1-bit raw AVI video

[Mats Peterson]

On 01/25/2016 01:26 AM, Michael Niedermayer wrote:
> On Wed, Jan 20, 2016 at 12:41:22PM +0100, Mats Peterson wrote:
>> I don't know about this one, since it adds some calculations inside
>> the loop, but it limits the line alignment to 16 bytes instead of 32
>> bytes as before. Less overhead.
>>
>> Mats
>>
>> --
>> Mats Peterson
>> http://matsp888.no-ip.org/~mats/
>
>>   raw.c    |    2 +-
>>   rawdec.c |   38 ++++++++++++++++++++++++--------------
>>   2 files changed, 25 insertions(+), 15 deletions(-)
>> 3c94aff2647b8a1b0e9c4c147b552c85255253ff  0001-v3.patch
>>  From 8d15b348fc84d46f7e764a2e9822aacc80e74407 Mon Sep 17 00:00:00 2001
>> From: Mats Peterson <matsp888 at yahoo.com>
>> Date: Wed, 20 Jan 2016 12:37:40 +0100
>> Subject: [PATCH v3] lavc/rawdec: Use AV_PIX_FMT_PAL8 for 1-bit raw AVI video
>>
>> The stuff about 1-bit video not necessarily being black & white in
>> QuickTime goes for AVI as well. Being 1 bit per pixel only means that
>> the data is bi-level. The two colors can be any color. Since many
>> 1 bpp AVI files don't have a palette following the BITMAPINFOHEADER,
>> I'm setting a "default" black & white palette in raw_init().
>
> Does the official player from microsoft play such crafted avi files
> with the palette colors or black and white ?
>
>
> [...]
>> @@ -94,8 +94,11 @@ static av_cold int raw_init_decoder(AVCodecContext *avctx)
>>               return AVERROR(ENOMEM);
>>           if (desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL)
>>               avpriv_set_systematic_pal2((uint32_t*)context->palette->data, avctx->pix_fmt);
>> -        else
>> +        else {
>>               memset(context->palette->data, 0, AVPALETTE_SIZE);
>> +            if (avctx->bits_per_coded_sample == 1)
>> +                memset(context->palette->data, 0xff, 4);
>> +        }
>
> hunk split out and applied
>
>
> [...]
>
>> @@ -222,18 +225,25 @@ static int raw_decode(AVCodecContext *avctx, void *data, int *got_frame,
>>               }
>>               linesize_align = 16;
>>           } else {
>> +            int j, row_pix = 0;
>>               av_assert0(avctx->bits_per_coded_sample == 1);
>> -            for (i = 0; 8 * i + 7 < buf_size && i<avpkt->size; i++) {
>> -                dst[8 * i + 0] = buf[i] >> 7 & 1;
>> -                dst[8 * i + 1] = buf[i] >> 6 & 1;
>> -                dst[8 * i + 2] = buf[i] >> 5 & 1;
>> -                dst[8 * i + 3] = buf[i] >> 4 & 1;
>> -                dst[8 * i + 4] = buf[i] >> 3 & 1;
>> -                dst[8 * i + 5] = buf[i] >> 2 & 1;
>> -                dst[8 * i + 6] = buf[i] >> 1 & 1;
>> -                dst[8 * i + 7] = buf[i]      & 1;
>> +            for (i = 0, j = 0; 8 * j + 7 < buf_size && i<avpkt->size; i++, j++) {
>> +                dst[8 * j + 0] = buf[i] >> 7 & 1;
>> +                dst[8 * j + 1] = buf[i] >> 6 & 1;
>> +                dst[8 * j + 2] = buf[i] >> 5 & 1;
>> +                dst[8 * j + 3] = buf[i] >> 4 & 1;
>> +                dst[8 * j + 4] = buf[i] >> 3 & 1;
>> +                dst[8 * j + 5] = buf[i] >> 2 & 1;
>> +                dst[8 * j + 6] = buf[i] >> 1 & 1;
>> +                dst[8 * j + 7] = buf[i]      & 1;
>> +                row_pix += 8;
>> +                if (row_pix >= avctx->width) {
>> +                    i += 4 - (i % 4) - 1;
>> +                    j += 2 - (j % 2) - 1;
>> +                    row_pix = 0;
>> +                }
>>               }
>> -            linesize_align = 32;
>> +            linesize_align = 16;
>>           }
>
> hunk split out and applied
>
> [...]
>
>
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>

Well, this one is still incorrect, since it doesn't cater for different 
strides.

Mats

-- 
Mats Peterson
http://matsp888.no-ip.org/~mats/