A Fast Hybrid Decision Algorithm for H.264/AVC Intra Prediction Based on Entropy Theory

Rate Distortion Optimization based spatial intra coding is a new feature of H.264/AVC standard. It efficiently improves the video coding performance by brutally utilizing variable block sizes and multiple prediction modes. Thus, extremely high computation complexity is required. This paper proposes a hybrid decision algorithm which can reduce unimportant block sizes and prediction modes. Entropy feature of each MB is extracted to decide the optimal block size. Then Prewitt operators based edge direction extraction module is designed to select only promising prediction modes. In addition, above two decision modules can function together to form a fast hybrid decision algorithm for intra prediction. Sufficient experiments prove that this hybrid algorithm is able to achieve 63% time reduction with PSNR loss limited to 0.07dB, bit rate increase at most 2.02%. It is expected to be utilized as a favorable accelerator for a real-time H.264 encoder for HDTV.