[FFmpeg-devel] [PATCH] Add DPX decoder

[Jimmy Christensen]

On 2009-05-06 15:44, Michael Niedermayer wrote:
> On Wed, May 06, 2009 at 02:35:30PM +0100, Jimmy Christensen wrote:
>
>>
>> Best Regards
>> Jimmy Christensen
>> Developer
>> Ghost A/S
>>
>> ----- "Michael Niedermayer"<michaelni at gmx.at>  wrote:
>>
>> | On Wed, May 06, 2009 at 01:56:56PM +0100, Jimmy Christensen wrote:
>> | [...]
>> |>  |
>> |>  | [...]
>> |>  |>  +	if(endian == 0x0) {
>> |>  |>  +		offset = AV_RB32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		bytestream_get_buffer(&headerBuffer, version, 8);
>> | headerBuffer +=
>> |>  | 8;
>> |>  |>  +		// Jump in extra 744 bytes to end in address 744 + 4 + 4 + 8
>> | =
>> |>  | 760 = 0x2f8
>> |>  |>  +		headerBuffer += 744;
>> |>  |>  +		orientation = AV_RB32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		w = AV_RB32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		h = AV_RB32(headerBuffer); headerBuffer += 4;
>> |>  |>  +	} else {
>> |>  |>  +		offset = AV_RL32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		bytestream_get_buffer(&headerBuffer, version, 8);
>> | headerBuffer +=
>> |>  | 8;
>> |>  |>  +		// Jump in extra 744 bytes to end in address 744 + 4 + 4 + 8
>> | =
>> |>  | 760 = 0x2f8
>> |>  |>  +		headerBuffer += 744;
>> |>  |>  +		orientation = AV_RL32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		w = AV_RL32(headerBuffer); headerBuffer += 4;
>> |>  |>  +		h = AV_RL32(headerBuffer); headerBuffer += 4;
>> |>  |>  +	}
>> |>  |
>> |>  | code duplication
>> |>  |
>> |>
>> |>  No, one is for big endian numbers, the other is for little endian.
>> | There are 2 types of DPX files.
>> |
>> | you duplicate a flexible endian 32bit reading function
>> |
>>
>> How does it work then?
>>
> static unsigned int read32(ptr, is_big)
> {
>      return is_big ? AV_RB32 : AV_RL32
> }
>
>

Thanks. :)

>> Sorry, I'm very new to ffmpeg and it's my first patch.
>>
>> |
>> |>
>> |>  |
>> |>  |>  +
>> |>  |>  +	s->width = w;
>> |>  |>  +	s->height = h;
>> |>  |>  +	s->bpp = 32;
>> |>  |>  +	avctx->pix_fmt = PIX_FMT_RGB32;
>> |>  |>  +
>> |>  |>  +	if(s->picture.data[0])
>> |>  |>  +		avctx->release_buffer(avctx,&s->picture);
>> |>  |>  +	if(avcodec_check_dimensions(avctx, w, h))
>> |>  |>  +		return -1;
>> |>  |>  +	if(w != avctx->width || h != avctx->height)
>> |>  |>  +		avcodec_set_dimensions(avctx, w, h);
>> |>  |>  +	if(avctx->get_buffer(avctx, p)<  0) {
>> |>  |>  +		av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
>> |>  |>  +		return -1;
>> |>  |>  +	}
>> |>  |>  +
>> |>  |>  +	buf += offset;
>> |>  |>  +
>> |>  |>  +	ptr = p->data[0];
>> |>  |>  +	stride = p->linesize[0];
>> |>  |>  +
>> |>  |>  +	for(x = 0; x<  s->height; x++) {
>> |>  |>  +		uint8_t *dst = ptr;
>> |>  |>  +		for(y = 0; y<  s->width; y++) {
>> |>  |>  +			rgbBuffer = AV_RB32(buf);
>> |>  |>  +			memcpy(&rgbField,&rgbBuffer, 4);
>> |>  |>  +			dst[0] = ((rgbField.R * 256) / 1024);
>> |>  |>  +			dst[1] = ((rgbField.G * 256) / 1024);
>> |>  |>  +			dst[2] = ((rgbField.B * 256) / 1024);
>> |>  |
>> |>  | discarding 2 bits in the decoder is unacceptable
>> |>  | also this way of reading is not portable
>> |>  |
>> |>
>> |>  How so? The last 2 bits are not used, since it's too small for alpha
>> | channel.
>> |
>> | ((rgbField.R * 256) / 1024);
>> | discards 2 bits
>> |
>>
>> You mean converts from 10bit to 8bit? Yes. But didn't know ffmpeg supported 10bit. Any suggestion on how to get the buffer 10bit is much appreciated.
>>
> using PIX_FMT_RGB48 would be an option
>

I tried using that but when trying to encode I get a "swScaler: Unknown 
format is not supported as input pixel format". Are there any codecs 
that use PIX_FMT_RGB48 which I can use for reference?


Btw. sorry about the breaking threads thing. Switched email client which 
should be better.

> [...]
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel