Fast Macro-block Selection Algorithm Using 2-D Haar Wavelet Features for H.264 Video Codec

Abstract

This paper proposes a fast macro block selection algorithm using Haar wavelet features for H.264/AVC video codec system. The proposed algorithm makes use of the two-dimensional Haar wavelet transform to estimate the sub-band energy of each macro-block in a given video frame. Then the sub-band energy can be applied to the macro-block mode decision as a primary parameter. To prevent the error segmentation, the proposed algorithm uses the rate-distortion cost (RDcost) mechanism in the mode decision procedure. Various experimental results show that the proposed algorithm can effectively make a macro-block mode decision under an RDcost value. Furthermore, the execution time of the proposed wavelet-based macro-block selection algorithm is faster 2~4 times than that of the high complexity mode algorithm with the similar macro-block mode decision results.