Dissertation > Industrial Technology > Radio electronics, telecommunications technology > Communicate > Image communication, multimedia communication > Image coding

Research and Implementation of Fast Motion Estimation Algorithm for H.264

Author ZhouDongHui
Tutor DuanMuChunJiang
School Zhejiang Normal University
Course Computer Software and Theory
Keywords H.264 Motion Estimation Mode Select Rate-distortion optimization
CLC TN919.81
Type Master's thesis
Year 2010
Downloads 90
Quotes 0
Download Dissertation

The latest international video coding standard H.264 is ITU-T Video Coding Experts Group (VCEG) and ISO / IEC Moving Picture Experts Group (MPEG) consisting of Joint Video Team (Joint Video Team, JVT) proposed in 2003. Since H.264 uses a number of new technologies, such as: a variable block size motion estimation (variable block size motion estimation), pel motion estimation (decimal pixel resolution motion estimation), multiple reference frame motion estimation (multi- reference frame motion estimation), rate distortion optimization (rate distortion optimization, RDO), CABAC (context-based adaptive binary arithmetic coding), CAVLC (context-adaptive variable length coding), etc., H.264 coding efficiency is superior to the previous video coding standard. However, also because the use of these techniques allows H.264 motion estimation of computational complexity is much higher than the previous coding standards. Because of this, reducing the complexity of the H.264 motion estimation algorithm of the recent international research hotspot. In full consideration of a variety of image sequences based on motion characteristics, we propose a vector-based prediction and multi-directional gradient descent search algorithm (Multi-Direction Gradient Descent Search, MDGDS) algorithms. The algorithm using the motion vectors of the time and spatial correlation on a search starting point to obtain the predicted vector, and then by using an adaptive threshold value to determine the type of motion of the current block, in order to use different search strategies. This algorithm can quickly expand your search and improve the search speed of motion estimation, thus avoiding to make the search process into local minima. Experimental results show that the multi-stage asymmetric cross hexagon search algorithm (UMHexagonS), simplified cross asymmetric multi-hexagon search algorithm (simplified UMHS) and enhanced prediction area search method (EPZS) traditional algorithm, The algorithm is able to maintain the accuracy of motion estimation, while further save a lot of coding time. By a plurality of coding results QCIF video sequence based on the statistical analysis, we found that the following three characteristics: the probability of each mode is adopted uniform; the pattern of neighboring macroblocks of the current macroblock mode there is a relatively large correlation; upper mode and the lower mode there is a certain correlation between. Meanwhile, the mode is not evenly distributed, and in some cases, we can ignore some of the coding efficiency and there is little chance limited mode. Based on these findings, we propose an approach based on statistical features for fast mode selection algorithm for H.264. The algorithm uses the coding mode to the early termination of the statistical characteristics of the model of the search motion estimation process, thereby reducing the computational complexity and improve the coding efficiency. Experimental results show that the high complexity mode (High complexity mode, HCM) algorithm, and fast, high-complex mode is selected (Fast high complexity mode, FHCM) algorithm, the proposed algorithm can remain nearly the same rate-distortion performance at the same Further to save a lot of coding time. The paper is organized as follows: First, the H.264 video coding principle, which focuses H.264 describes the principle of the motion estimation and key technologies. Next, a detailed description of JM15.1 (H.264 reference software) in an integer pixel fast motion estimation algorithm and fast mode selection algorithm. Then, focusing on what we proposed a new integer pixel fast motion estimation algorithm and a new fast mode selection algorithm.

Related Dissertations
More Dissertations