Advanced Techniques For High-efficiency Video Coding

Video-centric multimedia applications have penetrated into all areas of life.With the continuous improvement of resolution,high-definition video brings better visual experience and services to people,but at the same time it also brings a surge in data volume,causing storage and transmission problems,which also brings video encoding new challenges.The video-coding standard condenses the most efficient and the most advanced video coding technology of its time.H.265/HEVC(High Efficiency Video Coding)is currently the latest generation of international video coding standards.Similar to the previous generation of video coding standards H.264/AVC,H.265/HEVC still generally uses a hybrid video coding framework,but it is better in performance than H.264/AVC increased by about 50% BD-rate.Nonetheless,HEVC can continue to optimize to improve coding efficiency,and many studies have focused on how to optimize H.265/HEVC.The hybrid video coding framework consists of a number of key technologies,including transform coding,prediction technology,quantization,entropy coding,loop filtering,and rate control technology.In this paper,the transform coding and rate control technology have optimized separately,and a content-adaptive transform algorithm and a video coding and denoising joint optimization algorithm have designed to achieve efficient video coding.The introduction is as follows:1.The transform coding in the traditional video coding method uses a fixed transform basis function,and the residual image generated during the coding process displays non-uniform changes in content.This caused the inconsistency in the direction of the transform basis function and the residual image,which affected the efficiency of transform coding.In order to solve this problem,this paper proposes a content-adaptive transform.This paper analyzes the nature of orthogonal transform from the perspective of vector projection,and constructs a content-adaptive orthogonal transform based on the similarity between adjacent video images,and applied to the actual encoding.In addition,a variety of optimization method have designed in this paper to optimize and accelerate the selection of transform methods and the generation process of transform basis functions.After applying the new transform proposed in this paper to H.265/HEVC,it can significantly improve the coding efficiency in a specific coding mode.2.In order to achieve efficient encoding of noisy video,this paper uses H.265/HEVC's rate control mechanism to implement a joint video coding and denoising optimization algorithm.This method uses noise variance to establish a quantization parameter-estimation-model in image block level,and eliminates the influence of noise in the encoding process by assigning the optimal quantization parameter to each image block.On this basis,the H.265/HEVC rate control mechanism have further optimized,and a reasonable coding rate have assigned to the image to improve the "rate-distortion" optimization performance.The quantization parameter-estimation-model used in this method have obtained through offline training,so the increase in coding complexity is limited.At the same time,this method can realize video encoding and denoising simultaneously,which is more suitable for practical applications.