Reference Frame Recompression Algorithm Research And VISI Implementation For HEVC Encoder

Frame recompression is an efficient way to reduce the huge bandwidth of external memory for video encoder, especially for P/B frame compression. A novel algorithm, which is called In Block Prediction (IBP) reference frame recompression, is proposed in this paper. There are two subsets residual data produced by the proposed algorithm: predicted residual (PR, base layer) and truncated residual (TR, enhancement layer) which can provide different level reconstructed image——lossy and lossless. The lossy compression method is more efficient, the different reconstructed image is used in the different part of the HEVC encoder. The bandwidth reduction of the proposed method comes from two sources in our scheme, which differs it from previous designs and achieves much higher compression ratio. Firstly, it comes from pixel truncation. We use truncated pixels (PR) for integer motion estimation (IME), and acquire truncated residuals (TR) for factional motion estimation (FME) and motion compensation (MC). Since the pixel access of IME is much huge than FME and MC, it saves about 37.5% bandwidth under 3-bit truncation. Secondly, embedded compression of PR helps to further reduce data.The truncated pixels in first stage greatly help to achieve higher compression ratio than current designs. From our experiments,3-bit truncated PR can be compressed to 15.4% of original data size, while most of current embedded compressions can only achieve around 50%. PR is lossy data, which is generated by tail bit truncation (TBT) of original pixels, while TR is tail bits of original pixels. Combining PR and TR, it can easily reconstruct original lossless pixels. PR is then compressed by proposed in block prediction (IBP) and small value optimized variable length coding (SVO-VLC). The lossless compression ratio (PR+TR) can achieve 44.5%, which outperforms most of previous methods.According to our experiments, about 74.5% bandwidth can be reduced under 3-bit TBT. Moreover, IBP based IME saves about 35% of hardware cost. Our proposed algorithm is hardware/software friendly, and also fast IME algorithm friendly frame recompression scheme. Meanwhile the video quality degradation is negligible with small bit truncation.The main work of this paper includes:(1). Analysis and compare the existing algorithm, find their advantages and disadvantages.(2). Propose a frame recompression algorithm for HEVC which is hight-efficient and transparent.(3). Implement the algorithm with C++ and verify it.(4). Implement the frame recompression algorithm with very large scale integrated circuits (VLSI) and optimaize it. Simulate the frame recompression codec with EDA tools and synthesize it with Design Compiler.