Research On The Optimization Of High Efficiency Video Coding Algorithm Based On Region Of Interest

High Efficiency Video Coding(HEVC)algorithm is the latest generation of coding technology developed on the basis of H.264/AVC,which can reduce the code rates by more than 50%.However,the code rates reduced also lead to the increment of encoding complexity.As the increment of resolution and data in the videos,which results in more visual redundancy in the videos.The human visual attention mechanism is more conducive to filter out redundant information.Therefore,the research of high efficiency video coding algorithm based on Region of Interest(ROI)is more significant to reduce encoding complexity and improve the quality of output videos.The traditional Itti algorithm has some problems such as the lack of information in the ROI extracted,the inaccurate detection of ROI,and the positioning in edge of ROI is blurred.Therefore,in order to solve the problems existed in the traditional Itti algorithm,this paper improves the traditional Itti algorithm to extract ROI of videos.On the basis of the extraction of original brightness feature,color feature,and orientation feature based on the Itti model,this paper extracts texture feature and shape feature based on the spatial domain,which solves the problems that the lack of information in the ROI extracted and the positioning in edge of ROI is blurred after fusion,this paper extracts motion feature based on the time domain to solve the problem of the inaccurate detection of ROI.Then,under the operations of normalization and others,six single feature salient maps are generated:brightness single feature salient map,color single feature salient map,orientation single feature salient map,texture single feature salient map,shape single feature salient map,and motion single feature salient map.Finally,the information entropy fusion theory combined with the designed algorithm is used to adaptively obtain the weight of each single feature salient map,and under the cross-scale fusion operation,the extractions of ROI of videos are completed,which improves the accuracy of significant areas after fusion.Finally,the effective extractions of ROI of videos are achieved.In the intra prediction algorithm of HEVC,the quad-tree partition method of CU(Coding Unit)and the selection of 35 kinds of prediction modes of PU(Prediction Unit)increase unnecessary cost calculations of rate-distortion,which results in higher computational complexity in the intra prediction algorithm of HEVC.Therefore,in order to solve the problems existed in the intra prediction algorithm of HEVC,the designed algorithm based on ROI is used to optimize and improve the intra prediction algorithm.First,the designed depth decision algorithm of CU based on ROI with the idea of dimension reduction is used to determine the initial division depth of current CU.And the final division depth of current CU is determined by using the designed decision rule based on RONI,and the division of CU is terminated in advance to reduce the complexity of intra prediction algorithm.Secondly,two-dimensional discrete Fourier transform is introduced combined with the designed selection algorithm of PU based on ROI to determine the best prediction mode of current PU,which reduces the cost calculations of unnecessary rate distortion and reduces the complexity of intra prediction further.In the HEVC algorithm,the JCTVC-K0103 bit rate control algorithm has the problems of inaccurate bit weight estimated for the target bit allocation,ignoring the subjective observation mechanism of human eyes,and high calculation load in the parametric update stage.In view of the problems of the K0103 rate control algorithm,an optimized JCTVC-K0103 rate control algorithm based on region of interest is proposed.First,the bits allocated at the CTU layer are used as allocated weights for the complexity in the space-time domain to realize the redistribution of bit weights at the CTU layer.Then,the designed algorithm is used to allocate the weights of target bit allocation based on ROI once again,so that the output videos are more attuned with human visual attention mechanism.Finally,the quasi-Newton method is introduced to update the parametric model,which reduces the calculation load in the update process of parametric model.The proposed method in this paper is used to perform experimental comparative analysis based on the data provided by JCT-VC.The experimental results show that the method proposed in this paper can reduce the coding complexity,improve the output video quality and make the output video more in line with the human visual attention mechanism accurately and objectively.Achieving the optimization of high efficiency video coding algorithm based on region of interest,and which can provide some reference for the research work of high efficiency video coding algorithm.