Reasearch On Fast And High Efficient Algorithm For HEVC Lossless Compression

As the latest video compression standard,high efficiency video coding(HEVC)has been widely applied in people's work and life.Because lossless coding can maintain video quality after compression,it plays an important role in video surveillance,medical imaging,and autopilot.Because of the high complexity,the large amount of computation and the unsatisfactory compression effect of the compression process,the traditional HEVC lossless compression coding still has problems in practical applications such as low bandwidth,low latency and low overhead.In order to improve the compression rate and encoding speed,and reduce the transmission overhead of reference frame,this paper studies the intra-frame encoding,inter-frame motion estimation and reference frame compression algorithm in HEVC lossless compression.The main contents are as follows:Firstly,reaearch on DWT-based(Discrete Wavelet Transform-based)compression method for lossless HEVC intra coding.Aiming at the problem that the lossless intra-frame compression effect of HEVC is unsatisfactory due to the noncentralized distribution of pixels in spatial domain,this paper introduces DWT into the traditional HEVC lossless intra compression coding structure.The image is transformed from spatial domain to frequency domain,and then prediction and encoding for the frequency coefficients.After compressing the more concentrated high frequency coefficients,it will obtain much lower compression rate,lower CU splitting depth and less encoding computation.Experimental results show that,compared with the existing optimization methods,the DWT-introduced HEVC lossless intra compression can improve the compression rate and coding speed at the same time.Secondly,research on fast algorithm for intra coding unit splitting and mode selection.Aiming at the problem that the encoding time saving is limited by introducing DWT into HEVC lossless compression,this paper evaluates the texture direction and content complexity of the coding unit by texture detection method,and narrows the range of candidate for mode selection and CU splitting.Because of the centralized distribution of transform coefficients in high frequency sub-band region and the significant direction characteristics,detecting only four texture directions can accurately evaluate the variance direction and complexity of coefficients in the CU.Based on the results of texture and complexity evaluation,the prediction mode and splitting depth are filtered,which reduces the calculation amounts of encoding process.Experimental results show that combined with the fast CU splitting and prediction mode selection algorithm,DWT-based HEVC lossless intra compression can save nearly half of the encoding time.Thirdly,research on pixelwise adaptive prediction based lossless reference frame compression.In order to solve the problem that traditional reference frame compression algorithm can't predict independently for each pixel,this paper designs a pixelwise adaptive prediction mode selection algorithm for lossless reference frame compression.The algorithm estimates the similarity between current pixel and neighboring pixels based on the similarity of adjacent reconstructed samples.According to the similarity estimation,the most proper prediction mode is selected from horizontal,vertical and diagonal prediction modes for each pixel adaptively.Then,the predicted residuals are encoded by Huffman encoding process using the Huffman table designed in this paper to generate compressed bit stream.At the same time,to meet the random access requirement of reference frame compression,the size of encoding block is set as 32×32.Experimental results show that,compared with existing optimizations,the proposed algorithm achieves better reference frame compression efficiency with approximate algorithm complexity.Finally,research on adaptive search range adjustment algorithm based on motion vector prediction accuracy.Aiming at the motion estimation process which is the most time-consuming in inter-frame encoding,to narrow the search range and reduce encoding time of motion estimation,symmetric and asymmetric adaptive search range adjustment algorithms are proposed in this paper.According to the relationship between search range and motion vector prediction accuracy,motion vector difference is selected as the measurement of motion vector prediction accuracy.Based on the relationship between motion vector difference and block matching error multiples,we obtain the estimated value of motion vector difference by approximately calculating the matching error multiples utilizing the block mathching error of upper depth CU as the estimation of current block matching error.The symmetric adjustment algorithm utilized the estimated motion vector difference as the search range for horizontal and vertical directions.However,in order to adapt to the different motion intensity of objects in horizontal and vertical directions,the asymmetric algorithm refers to the adjacent encoded CUs.Based on the similarity of motion and texture,combined with the motion vector prediction accuracy,we define the similarity and extension coefficients to calculate the compuatation weights.The search range of horizontal and vertical directions are adjusted independently utilizing the motion vector differences of adjacent CUs and computation weights.For the case of adjacent CUs missing,the symmetric search range adjustment is used to compensate.Finally,the encoding performance of the proposed algorithm is evaluated.Experimental results show that the proposed algorithm can accurately estimate the search range of the prediction unit and significantly reduce the encoding time without notable deterioration in bit-rate.This paper optimized the lossless HEVC intra-frame coding,reference frame compression and motion estimation of inter-frame process,enhanced the compression ratio of intra-frame encoding,saved the encoding time of intra-and inter-frame compression,reduced the transmission and strorage cost during motion estimation and increased the coding efficiency of HEVC lossless compression.