| This research proposes a block-based video codec design with two objectives. The first goal is to propose a method for intraframe that improves the rate-distortion (peak signal-to-noise ratio versus bit rate) of a fixed-size transform encoder. The proposed method uses three integer transform sizes (4×4), (8×8), and (16×16). The codec also adopts H.264-like spatial prediction to intraframe encoding. For simplicity of the design, Huffman variable-length code is used as entropy encoding. For intraframe encoding, the simulations show rate-distortion improvement over JPEG and JPEG2000. In some test sequences, the simulations also show improvement over H.264 (baseline profile at low complexity mode without rate-distortion optimization) with a small increase of operations on each macroblock at the decoder side.;The second goal of this research is to study rate-distortion behavior of the interframe codec with novel motion estimation based on structural similarity (SSIM) and the codec with conventional motion estimation based on pixel error distortion (sum of absolute difference). A study from previous literature shows that the structural similarity metric provides better image assessment than a pixel error based metric (mean square error and peak signal-to-noise ratio). Structural similarity measurement on the true color components (RGB) with equal weight for each component is proposed. The results on rate-distortion show that both structural similarity and peak signal-to-noise ratio (PSNR) provide similar measurements. Both sum of absolute difference (SAD)- and structural similarity (SSIM)-based distortions in motion prediction of large block sizes, {(16×16), (8×8)}, have similar performances. For the small block size of (4×4), SAD-based distortion provides better rate-distortion performance. Distortion calculation for SSIM requires more operations compared to SAD. |
