Screen Content Coding Optimization Based On Content Characteristics

With the rapid development of the Internet and cloud computing,screen content video(SCV)is widely used in multimedia applications such as mobile devices,remote desktops,virtual desktops,and wireless displays.Efficient compression of SCV has become very important in order to meet the bandwidth requirements.At present,International Organization for Standardization has formulated screen content coding(SCC)based on high efficiency video coding(HEVC)according to the content characteristics of screen content video.Compared with camera-captured content,screen content frequently contains no sensor noise,and such content may have large uniformly flat areas,repeated patterns,highly saturated or a limited number of different colors,and numerically identical blocks or regions among a sequence of pictures.The screen content coding(SCC)added coding tools such as intra block copy(IBC),palette mode(PLT),adaptive color transform(ACT)and adaptive motion vector resolution(AMVR)to improve coding efficiency.Due to the limited coding resources of real-time systems,however,the encode complexity and resource allocation of the SCC need to be optimized.In this paper,the complexity optimization and rate control optimization of screen content video coding are investigated as follows.(1)In this paper,we propose a fast intra coding unit(CU)depth decision based on ensemble learning.The depth decision problem in CU is regarded as a three-layer two-class problem.Firstly,the L1-loss based linear support vector machine(SVM)is employed as basic classifier for its simplicity.Then,a bagging scheme is applied to train the linear classifiers and boost the prediction accuracy by ensemble learning.Compared with the reference software SCM-5.0,the proposed scheme can achieve 30% complexity reduction on average with only 1.64%-bit rates increase.Besides,we also propose a fast decision algorithm for predictive unit prediction mode based on optimal stopping theory.The algorithm divides the mode into four classes,including Intra,IBC-Merge,IBC and PLT,according to the traversal process of the prediction unit mode and the proportion of the mode as the best mode.Then,the proposed algorithm only checks the mode before the optimal stopping point to reduce the calculation amount of the mode traversal.In addition,this paper studies the relationship between the coding information feature and the best mode,and then estimated the probabilities of Intra,IBC-Merge,IBC,PLT,by multi-class classifier.Experimental results show that the proposed algorithm can save 11% of the coding time on average.(2)In the view of rate control model used in SCC,this paper explores the relationship between SCV image features,quantization parameters and targeted bit rate,and proposes an initial quantization parameter prediction algorithm based on support vector regression.The algorithm uses the statistics of energy,entropy,contrast and correlation of the targeted bit rate,and gray level co-occurrence matrix as input to predict the quantization parameters of the initial frame.Compared with the SCM8.0,the algorithm proposed in this paper increases the average BDPSNR by 0.47 dB at the same bandwidth.Besides,we proposed a two-pass frame level rate allocation algorithm for screen video,in which the current GOP is pre-encoded for coding information collection.In the second pass,the weight factors of each frame in the current GOP is updated based on the analysis results from the first pass.Experimental results demonstrate that our proposed algorithm can achieve better RDO performance for sequences with fast scene changes.