vcodec_dct
is a toy codec based on ideas of AVC (H.264) standard, but with lots of simplifications, luma only,
because it is developed just for fun.
As of now, only intra-frame compression is implemented, featuring 8x8 integer transform (aka simplified DCT used in H.264) with 4x4 Hadamard transform for DC coefficients and 3 spatial prediction modes. The inter-frame compression (mostly motion prediction) is WIP, and I hope to finish it some time later.
Encoding:
./vcodec-test /path/to/Y4M-luma-only-raw-video > /path/to/encoded-output
Decoding:
./vcodec-dec-test /path/to/encoded-file > /path/to/decoded-y4m
You can play Y4M files with ffplay
, for example.
Compression ratio/PSNR are still to bad to brag about it.