Research And Improvement On Multi-Reference Frame Selection And All-Zero Block Detection Algorithms For H.264

With the rapid development of information technology and internet technology in the 21st century, multimedia information, especially video information, begins to take place of the plain text information gradually and become the main carrier from which people acquire knowledge. Applications such as video conference, video phone, distance learning and training, telemedicine, video-on-demand (VOD) have gradually entered into people's daily life.However, uncompressed video data occupies quantity of storage space and transmission bandwidth for its high redundancy characteristic; and using it directly is neither economic nor practical. The emergence of video codec technology which reduces the cost of using video signal provides a good solution to this problem.As a next-generation video standard, H.264 has aroused people's extensive attention from the day of its birth. It has higher coding efficiency and better adaptability of the network than existing video standards which attracts more and more researchers to study on it. However, as the introduction of many new technologies such as variable block-size modes, multiple reference frames selection, fraction-pixel motion estimation and so on, the encoding complexity of H.264 has been increased heavily compared with the previous video standards. It is the main reason that H.264 is hard to be used in real-time applications. As a result, reducing algorithm complexity, accelerating encoding speed, and meeting requirements of real-time network video transmission have become a hotspot of current researches. In this article, we firstly introduce the background of this thesis, as well as the present status of research on video coding algorithms; and then introduce the H.264 video standard briefly; finally, choose multi-frame selection and all-zero block detection as the two main aspects for optimizing H.264 coding algorithm, and validate its correctness experimentally.Motion compensation based on multi-reference frames is the core of H.264, and it is also the most time-consuming part. Through research we find the impact of temporal and spatial motion correlation on reference frame selection is distinct for different block size modes. In this paper, we study the statistical properties of video sequences and the spatial correlation in video frames first, and then propose an adaptive fast multiple reference frames selection algorithm based on motion correlation. It can take out the reference frames that are not necessary for motion compensation, and reduce the complexity of coding adaptively according to the features of the video sequences. The experimental results show that our algorithm can save more than 40% of coding time on average with negligible loss.The residual signal gained after motion estimation needs to be transformed and quantized prior to entropy coding in H.264. In many cases, especially in low bitrate situation, the probability we obtain all zero quantized block (AZB) is high. In this paper, we study some of the existing fast AZB detection algorithms, then we use divide and rule strategy inversing the formulae of integer transform and quantization after we study the characteristics of 4x4 integer transform. By this way, we get a new sufficient condition for AZB detection. The experiment result shows our algorithm can find out more AZBs compared with existing methods without any loss and save more computations at the same time.