Research On High Performance Intra Video Coding

With the fast development of the internet and mobile terminals, andthe improvement of cloud computing, big data technology, multimediaapplications and services have become an important part of people s dailylife. As the core of multimedia communication technology, image andvideo coding technique has always been a hot topic in the field ofinformation processing. Besides, the demands of video dimensions from2-D to3-D and video resolutionsfrom HD to ultra HDlead to higherrequirements for more efficiency video coding technology. In order tobetter adapt to the requirements, the latest video compression standardHEVC with higher compression efficiencyemerged. HEVC has a nearlydouble coding efficiency compared to H.264/AVC.The coding framework of HEVC is still inherited from H.264/AVC,including prediction, transform and entropy coding.It is the techniqueinnovation of these modules under the previous framework that improvesthe coding efficiency. For intra coding, two problems arise because ofthe improvement of video resolution.One is that smaller size of thecoding unit is preferred due to the complex structures of the picture.However, this will introduce more bits to represent the intra modeinformation.The other problem is that, if the size of coding unit is toobig,the distance between the reference pixels and the pixels to bepredicted will increase and the prediction accuracy willreduce..H.264/AVC video coding standard uses coding blocks of fixedsize and seems to be inflexible. Considering this deficiency, HEVCadopts a new conceptual coding tree unit to obtain a coding unit ofvarious sizes. What s more, more intra prediction modes are utilized toprovide a more sophisticated intra prediction.HEVC intra prediction has35different modes, including Angularprediction modes, which include33directions, planar mode and DCmode. In this paper, we give a detailed analysis of the key technicalpoints related to the intracoding and down-sampling technology, sum up the existing problems as well as some important improvements, and anovel intra prediction method based on node-cell structure prediction isproposed in this paper.In Node-cell structure, pixels in the current coding unit are dividedinto two sub-sets, which are node pixels and pixel cell pixels, respectively.They are predicted and encoded separately.Node pixels can be regardedas down-sampling from the block pixels, Except the node pixels, the restpixels in the block are called cell pixels.Intra prediction based onnode-cell structure includes planar mode, DC mode and Angularprediction under different down-sampling rate. For Angular prediction ofnode-cell structure, node pixels reconstruction can be used for cell pixels prediction, whichcan achieve bi-directional prediction and higherprediction accuracy and further improvement of the coding efficiency.Toensure the quality of reconstruction, the residuals of node pixels and cellpixels are transmitted to the decoder.Experimental results show thatcompared with the intra coding of HEVC, the intra coding based onnode-cell structure improves the intra coding performance.On the basis of the intra prediction algorithm of the node-cellstructure, a quantization parameter control algorithm is proposed. Thealgorithm considers that the node pixels and cell pixels are encodedseparately and can be reflected by each other. According to the features ofdifferent coding units, optimal transform coding quantization parametersof node pixelsare adjusted. Experimental results show that this algorithmfurther improves the encoding efficiency without increasing the encodingtime and computational complexities, compared with the intra codingalgorithm based on the node-cell coding structure.