I implemented a 32-bit float GBRAP/GBRP pixel format and used it in vf_zscale.c and vf_overlay.c. I also implemented an EXR encoder using miniexr that uses the new float pixel format. I can contribute this if there is interest. I used the above mods to do HDR video processing that includes overlaying HDR and SDR video. IMHO, for ffmpeg to handle HDR video well, float and half-float are needed and also metadata to signal the color space and transfer function for each picture. It's a big job that can touch a lot of code, which is why I kept my changes private and limited to only what I need.
From: Martin Vignali <martin.vign...@gmail.com> To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org> Sent: Tuesday, March 7, 2017 4:16 PM Subject: Re: [FFmpeg-devel] libavcodec/psd : add support for rgb/gray float Hello @Carl : I make some tests in order to avoid temp buffer, but i have some strange result, when i not use a temp buffer. > Time to think about a float pixfmt? I find it weird that we support > excessively obscure pixfmts like AV_PIX_FMT_BGR4_BYTE, but then have > hacks like this one if the codec uses float data. > > (I'm just commenting, not arguing for/against your actual patch.) > I agree with your comment. Float pixfmt is a more clean way. it can also be useful for image processing with filters, and encoding float data. Any tips to add support for float pixfmts ? I try to take a look to the source code, but it's not very easy to understand. Martin _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel