Research Of Scalable Video Coding Based On H.264/AVC Extension

With the rapid development of our society, the communication ways is becoming more and more. We also have higher requirement of the communication. So, the application of multimedia in daily life is becoming important increasingly. In addition, we give most hope to video communication in various multimedia ways. But different terminals and customers have different requirement of the video. This is the basis of the scalable video coding. The essential theory of SVC is that making sure part of the basic video data can transmit safely under the limited network condition and the enhancement data only transmit under the good condition. In 2004, the SVC extension part of H.264/AVC was first becoming a formal standard. During the coding process, it almost has every algorithm in H.264/AVC and also has its own inter-prediction. Because the data in enhancement layer is coming from the basic layer, they have strong correlation. The inter-layer prediction can eliminate the correlate between the layers.However, the algorithms used in standard H.264/AVC have higher complexity. If we consider inter-layer prediction in its SVC extension, the complexity will become unacceptable. The coding process will cost a lot of time, and it can not be used in daily life. So, the optimization of the SVC extension is the most important thing. Some experimental results show that the mode decision in the coding process cost most time. There were many fast mode decision methods before. They nearly all use the corresponding MB in base layer to predict the mode in enhancement layer. The experiment shows except for the corresponding MB, the neighboring MBs surrounding it also have a considerable effect on the mode decision in enhancement layer. So, the paper propose an method take into account the neighboring effect of each MB, to reduce the candidate mode number and the mode decision process will become more fast.Except for the above fast mode decision method, the paper also propose the other method for the coding process optimization—optimization based on the ROI. We choose the important MBs in the image according to the characteristic of HVS(Human Visual System). Then, we only give these MBs enhancement layer coding. Before the enhancement layer coding, we first compare the weight of the current MB with the valve value computed by the FJND model. If the weight is bigger than the valve, the MB will be coded in enhancement layer, because the MB is the visual interesting region. We use this method not only can reduce the coding time, but also get a good subjective quality of the video sequence. We use visual subjective sense to optimize the process of video coding.