The Research On FGS Video Coding Algorithm Based On HEVC

With the development of network, mobile communication and video application,the requirements of video coding is higher and higher, not only the highercompression of video coding technology, but the good adaptability of bitstream aftercoding. Among them, the fluctuation of network bandwidth is a fundamental problemneed to be solved. The traditional method used to solve the problem are Broadcastand Streaming-conversion, however, the bitrate is restricted in specific bitrates beenset in advance and difficult to dynamic switching. The fine granularity scalable videocoding technology been put forward in MPEG-4, it can match the fluctuation ofcurrent bandwidth finely, but compared with the non-scalable coding efficiencyreduces2-3dB. With the development and perfection of H.264/AVC, scholarsproposed the fine granularity scalable video coding technology which is based onH.264and a series of improvement, all these improve coding efficiency to somedegree, but the coding efficiency is still a little low.New generation video coding standard—High Efficiency Video Coding(HEVC),has double efficiency of compression than H.264/AVC, that is to achieve50%bitratesdeceasing and keeping the same video quality at the same time. In order to improvethe efficiency of the fine granularity scalable video coding, this paper combine HEVCand FGS, propose a FGS video coding scheme which is based on HEVC, and basedon this, propose three improvement schemes. The FGS which is based on HEVCadopts the HEVC coder in base layer to improve coding efficiency, uses MPEG-4FGS enhance layer in enhance layer. The enhance layer using bit-plane codingtechnology and the stream of the enhance layer can be truncated at any position, thusable to match the change of network bandwidth. The former improvement algorithmutilizes the segmentation of HEVC coding unit. The sports vigorous area adoptscoding unit division of small size and flat area use larger size generally. Throughstatistics the segmentation of coding unit, adaptive find details in the image as ROI,raise bit-plane, priority coding and transmission. When the stream of enhance layer istruncated, the information of ROI can be preserved preferentially and the subjective effect can be improved greatly. The second improvement algorithm takes theadvantage of OpenCV in video images, detect the moving objects and set it as ROI,raise bit-plane, priority coding and transmission, the subjective effect can beimproved greatly. The third improvement algorithm adopts motion compensation loopto enhance layer to improve coding efficiency of enhance layer by removing thetemporal relativity of original video in enhance layer. The motion compensation loopof enhance layer is combined with the base layer to reduce the time, so that thecoding unit in enhance layer is no longer divided, and the motion vector and referenceframe indexes of prediction unit in base layer are tranfered to enhance layer directlyto reduce the encoding time. Introduce the enhance layer information which candecode correctly to base layer to increase the coding efficiency by using the higherquality reference frames. As to the problem of prediction drift, this paper adapts theadaptive front network estimation algorithm make sure the number of bit plane whichcan correctly transmits.