Research On Scalable Video Coding

With the development of Internet and wireless network, a great deal of multimedia data, especially video data, need to be transmitted over network. It makes the goal of video and image coding changed. Traditional video coding is for storage. The major task of the video coding techniques for storage-based applications is how to improve the coding efficiency. However, transmitting digital video over network has to deal with two major problems: bandwidth fluctuation and packet errors. So the goal of video coding is to adapt to network bandwidth variations and tolerate transmitted errors besides the requirements in higher coding efficiency. To address these issuses, scalable video coding technique has been developed. It allows compressing the video data once and then decompressing it at multiple data rates, frames rates, spatial resolutions, and/or video quality (SNR). Such a compression technique would be very desirable from a networking viewpoint as it allows differentiated quality and bit rates depending on the kind of service chosen by the user. There has been a great deal of work in scalable video coding. But highly efficient scalable video compression algorithms have not existed. Error tolerance is not considered in existed scalable coding methods. In this paper, we will focus on the challenges and issues involved in scalable video coding. The creativities and contributions are discussed in detail as follows:1. Video transmission over network channels often meets packet loss and errors which deteriorate the quality of reconstructed image. Multiple description coding (MDC) is a new approach to alleviate the problem of packet loss in image communication over erasure channels, such as Internet and wireless networks. But existed MDC algorithms produce no scalable bit streams. It just enhanced error tolerance and has not bandwidth adaptability. In this part, combining MDC and fine granularity scalable (FGS) coding method, a multiple description scalable coding method is proposed. It produced spatial layered and SNR FGS bit streams with strong robustness. There are three main contributions as follows:(1) To enhance robustness of compressed video to packet loss, a multiple description algorithm based on polyphase down sampling is proposed in this thesis. Input video is split into 5 descriptions with different image resolution. Three motion estimation methods are used for these descriptions. Accuracy of motion estimation is improved. Motion vectors are coded in multi-descriptions, which also enhanced robustness of streams. (2) Based on above multiple description algorithm, a multiple description scalable coding method (MDSC) is proposed to improve the bandwidth adaptability of bit streams. Five descriptions are separated into tow base layers and three enhanced layers. The enhanced layers are coded with FGS. Experimental results show that MDSC enhanced robustness and bandwidth adaptability of bit streams.(3) The MDSC coder is implemented based on FPGA and DSP. The preprocess parts based on FPGA is expounded in details.2. For providing interactive and scalable video services, a content-based MDSC method is proposed based on MPEG-4 FGS. The idea of MDSC is adopted in this method. Segment video images into VO (video object), and then MDC method is used to code important VO such as motion objects. In addition, for video monitoring system, a non-linear polyphase downsampling algorithm is used in MDSC system. Content-based MDSC method enhanced robustness of the interesting part of the image.3. An image quality enhancement video coding method for interesting regions is proposed. Segment interesting regions from the scene by moving object detection. An adaptive motion estimation algorithm named RAME is proposed in this paper, which has improved the accuracy of the prediction. RAME and the variable-size-block binDCT are used in motion regions. It has improved the image quality in these regions. On the basis of the traditional temporal scalable coding, a progressive temporal scalable coding method is proposed, which strengthened the bandwidth adaptability of bit streams by using granule temporal scalability, and adopt enhanced layer as predict reference to improve accuracy of motion estimation. Combining the image quality enhancement coding with the progressive temporal scalable coding, the codec system model is presented, which uses different coding scheme for motion regions and the background region. The simulation results show that this method improved the whole reconstructed image quality, and made local image quality of the motion region topper and smoother than other regions.4. An adaptive and scalable video coding method based on shot segmentation is proposed, and is realized for distance education application. An algorithm for video shot segmentation (VSSB) is proposed. The complexity of the algorithm is logarithmic. The algorithm constructs a binary tree with bisearch, and finds shot segmentation points with the breadth-first search. For real-time application, a shot segmentation method with two thresholds is used. Based on shot segmentation, a codec system model is established. In this codec system, a parallel data compression approach is adopted for decreasing the compression time. Experiment results show that the video coding algorithm based on short segmentation increased the compression ratio and raised the compression efficiency. In addition, color histogram is used, which segment video image of distance education into two categories. For these two sets of images, different coding method and scalable approach are used.