Research On Predictive Coding Method Of Screen Teaching Video Using Markov

With the development of modern distance education technology,screen teaching video is widely used in high-quality college courses,Internet distance education,and cloud open classes.However,the traditional video encoder can not meet well the needs of the off-line storage and online transmission of massive high-definition teaching videoes.Therefore,it is much important to research a coding method suitable for the content of screen teaching,which is also needed by the modern distance education.First,the paper analyzes the importance of screen teaching video to modern distance education,and discusses the need of screen teaching video compression.Second,on introducing the data statistics of the screen content,this study discusses the development of screen content coding methods.State-of-the-art methods are classified into six classes,namely palette-index map based methods,template matching based methods,block matching based methods,dictionary-based methods,shape representation based methods,as well as chroma component coding methods.This study comparess,analyzes,discusses the advantages vs disadvantages,as well as the application scope of various methods.Based on the above,this study carries out the research on the prediction algorithm of palette-index map.Third,the study analyzes the main reason of prediction failure of both the multi-stage prediction(MSP)method and the local directional prediction method presented in the edge gradient areas or conjunction areas.Subsequently,this study describes the local correlation between adjacent indexes using a 1-order 2D Markov stochastic process model and a color transition probability,and proposes a local prediction algorithm for the index map based on the Markov model.Experimental results showed that the prediction accuracy of the proposed algorithm reached 97.53%,which was on average 4.33%and 2.10%higher than that of the MSP method and the local directional prediction method,respectively.Fourth,aiming at the indexes suffering from local prediction failure,this study further addresses its non-local prediction strategy.After analyzing the main reason of prediction inefficiency of typical 4-neighborhood template,this study designes a 2-neighborhood matching template containing 4 sub-templates,and models the template prediction using a transition probability that can be implemented by table lookup operations.Subsequently,this study puts forward a non-local prediction algorithm for the index map based on 2-neighborhood Markov joint transition probability.Experimental results showed that the prediction accuracy of the proposed algorithm reached 97.70%,which was a 4.49%and 2.27%on average increase over the MSP method and the local directional prediction method,respectively.Finally,the work of this study is summarized and forecasted.