Comparisons of Three Video Coders

The video coders of V-SPIHT, MC-EZBC, and H.264/AVC were tested on the QCIF (Y 176x144, U/V 88x72) sequences "carphone" and "mother_and_daughter". The conditions were 16 frame GOP's for V-SPIHT and MC-EZBC, and 8 frame GOP's for H.264. The frame rate for all is 30 frames/second.

V-SPIHT is a scalable wavelet codec with simple 1 pixel resolution motion compensation. From one encoded file, any decoding rate can be set almost exactly.
Both the encoder and decoder run in real-time in software.

MC-EZBC is a scalable wavelet codec with complex 1/4 pixel resolution motion compensation. From one encoded file, any decoding rate can be set almost exactly.

ITU standard H.264/AVC is a non-scalable DCT codec with more complex 1/4 pixel resolution motion compensation. Target rates can not be set. A quantization parameter (QP) is set to obtain an encoder rate which can not be known beforehand. New encoder and decoder runs must be made for every rate.

The PSNR results given are average PSNR in dB per frame.

Click on the links directly below to see graphs of the results.