Abstract:

The suitable modeclass is adaptively selected for the current macroblock (MB) based on the context of encoded MBmodes in the spatiotemporal neighborhoods of the previous and current H264 video frames. Two matrixes which are composed of sum of absolute values of the integer transform coefficients in 4×4residualblock contained in luminance 16×16block and 8×8block are defined respectively, the novel MB and subMB mode selection criteria which use the derived gradients of the two matrixes are then proposed. Experiments show that the proposed algorithm can averagely save about 11% and 7% encoding time compared to joint video team (JVT)J033 and JVTK021 respectively with almost no rate distortion efficiency degradation.

Key words: video coding, rate distortion optimization, mode selection