[FFmpeg-devel] [PATCH]Fix pnm encoding with bpc set

[Reimar Döffinger]

On 25 August 2014 06:41:58 CEST, "Reimar Döffinger" <Reimar.Doeffinger at gmx.de> wrote:
>On 24.08.2014, at 14:02, Christophe Gisquet
><christophe.gisquet at gmail.com> wrote:
>> Hi,
>> 
>> I think a first point must be cleared, seeing your reaction to my
>> pnmdec comment. Here' are the opinions:
>> - Mine: bits_per_raw_sample should indicate the dynamics of the
>signal
>> when it is different from the colorspace bitdepth
>> - ffmpeg's and yours (if I'm not mistaken): the signal bitdepth
>should
>> always match the colorspace one, and bits_per_raw_sample should
>> indicate the original bitdepth
>
>The colorspace is supposed to be authoritative.
>If it says 16 bit, then showing it as 16 bit must work properly.
>The only intention of bits_per_raw_sample is to indicate that there is
>no point in storing some of the lower bits if you have the option.
>
>> I believe rescaling in the decoder, like pnmdec does, is wrong. Like
>> CMYK support, doing that is the job of a rescaler independent of the
>> codec (yes that would be a speed hit). Otherwise, several "codecs"
>> (dpx comes to mind) would have to implement it, up to getting it
>> slightly wrong.
>
>That can be solved by a common function if necessary, alternatively a
>12 bit colorspace could be added, but the cost/benefit is questionable.

Also note that it is perfectly possible to check this is handled correctly, though how to best do so may be up to dispute.
For all components,
(c & ((1<< (16-rawbits))-1)) == (c >> rawbits)
Must hold.