Research On Screen Content Coding Based On HEVC/H.265

Screen Content refers to image/video generated by electronic devices, like cartoons,3-D games or captures of computer screens or video with text overlay etc. With the development of Mobile Internet and cloud computing, screen content coding has attracted more attention. Current video coding technology based on prediction and transformation such as HEVC can’t compress screen content efficiently. The display principle and characteristic of screen content as well as HEVC compression influence is analyzed in this paper. A pattern matching mode is presented at the aim of improving the coding efficiency of screen content on the basis of HEVC coding framework.The research work of the paper is detailed as follows.(1) The display principle of original screen content and video application’s screen content is analyzed. Furthermore, the property of screen content is analyzed with the comparison of natural content. In addition, the influence of compressing screen content with HEVC coding framework is analyzed. Finally experiment is conducted by using HEVC to compress natural video and screen content, in order to evaluate that HEVC coding framework based on prediction and transformation doesn’t perform well on screen content.(2)A pattern matching mode is presented as an additional coding tool for improving coding efficiency of screen content, which will be incorporated into HEVC coding framework. Firstly color preprocessing is conducted to screen content in the mode, which will turn YUV pixel value into basic color and index map, i.e., color component and structure component. Different coding methods are adopted because color component and structure component of screen content have different characteristics. Predictive coding is conducted to basic color, and the prediction result is coded in a two-level hierarchical coding scheme. Explicit differential encoding is applied to the basic color which isn’t predicted successfully. Dictionary coding method is performed to the index value in order to take advantage of repeated pattern and reduce the redundancy of structure component. String search is conducted on the dictionary to find whether there is the string matching with the current index string to be encoded. If the matching string exists, current index string to be encoded will be encoded as the offset and length of continuous matching, otherwise predictive coding will be adopted.(3)It’s required to find the longest matching string in the dictionary when the dictionary coding is conducted to index value in pattern matching mode. String matching operation may be performed many times while the operation is time-consuming. A fast dictionary searching algorithm based on hash-table is proposed to reduce encoding time, which will decrease the times of string matching operation significantly. New syntax elements and entropy encoding process are defined since different encoding framework is adopted in pattern matching mode compared with HEVC. In addition, an encoding scheme of matching string similar to k-order exponential Golomb coding is proposed since the probability distribution of matching length isn’t equal in the dictionary coding of index value.(4) A method for screen content capturing is proposed based on FFMPEG, which can store the original screen content losslessly in the YUV file with specified frame rate. Experiment is performed to validate the code efficiency of the proposed pattern matching mode on the basis of HEVC.Finally both the pattern matching mode and HM are employed to compress screen content which is recorded using the above method for screen content capturing. The subjective quality assessment is performed on the decoding picture.