[FFmpeg-devel] [PATCH]Fix pix_fmt detection in jpeg2000

[Paul B Mahol]

On 7/17/13, Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:
> Nicolas Bertrand <nicoinattendu <at> gmail.com> writes:
>
>> Le 17/07/2013 11:19, Carl Eugen Hoyos a ecrit :
>> > But imo, the important question are not valid
>> > samples but why do you think that the decoder
>> > should depend with its MCT decision on the
>> > colour space (especially if the specification
>> > does not mention colour space at all).
>>
>> Because, as MCT is defined, if the input color
>> space is YUV, there are no need to do the MCT.
>
> Isn't this completely up to the encoder to decide?
> (Unrelated to the fact that it probably makes no
> sense.)
>
>> If the MCT is done with a YUV color space, we
>> will do YUV --> RGB --> YUV. so useless.
>
> Or in other words:
> If the encoder decided (because of an error or
> another reason) to run its yuv input through
> the RGB -> YUV conversion, wouldn't the MCT be
> necessary to produce correct output?

Pointless discussion asside, does yuv samples (subsampled or not)
decode anything
sane with this patch?

>
> Carl Eugen
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>