Project meeting this week was cancelled.
I have implemented video decoding and encoding in C/C++ using ffmpeg. I can now encode the sequence (337MB) to a video file (26.6MB), and then decode it back to the exactly same sequence, losslessly, using h264 with yuv444p pixel format.
The decoding program can accept any video formats ffmpeg supports. Therefore, I can post-process/transcode the video using any existing programs, for example, the ffmpeg command line tool. The decoding program will still be able to read the video.
I have tested a few other encoding codecs and quality parameters by transcoding, and tested pixel accuracy using ffmpeg’s PSNR filter, the results are shown below: (PSNR in dB)
Input (original) formats: yuv444p_yuv: Channels ordered as packed YUV (yuvyuv...), encoded as yuv444p yuv444p_y-u-v: Channels ordered as planar YUV (yy...uu...vv...), encoded as yuv444p yuv444p_rgb: Channels ordered as packed RGB (rgbrgb...), encoded as yuv444p rgb24_rgb: Channels ordered as packed RGB, encoded as rgb24
Transcoding output formats: [encoding]-[quality factor (lower the better)]
Uploaded some sample videos, with 8x upscaling: Vixen Project playlist However, the video player uses interpolation, so the video will be distorted. The original video has clear boundaries between pixels.
I will be working on porting encoding & decoding to C#, recording controller information to media file metadata, and mux audio & video channels.