在可视IP电话、终端会议等多媒体传输和存储中，由于表示视频信息的数字需要大量的比特数，对信道带宽和存储空间提出了很高的要求，视频压缩便成了解决此问题的有效途径。而运动估计技术是视频压缩的关键技术，它是一种帧间编码的方法，其基本原理是利用帧间的时间相关性，减少时间冗余度。一种优秀的运动估计技术就是如何得到精确的运动矢量，通过用当前部分在相邻帧中寻找最相似的部分作为匹配部分，用运动矢量指向匹配部分，再用匹配部分与当前部分之间的差形成误差部分，以匹配部分和误差部分来重建高质量的压缩视频图像。为此，以运动估计技术为核心的视频压缩编码技术中，搜索运动矢量的速度和运动矢量的准确度是影响视频重建的关键因素。 本课题在总结已有算法的基础上，重点研究运动估计技术存在的问题，通过实验仿真，研究了全搜索算法、三步搜索算法和菱形搜索算法，详细分析了它们的优缺点，为进一步研究提供科学的理论和实验数据。根据视频图像质量和算法运行时间两者之间的关系以及运动矢量的时空相关性和中心偏向特性，提出了预测三步搜索算法、预测菱形搜索算法以及自适应菱形搜索算法，并对各种算法进行了仿真。结果表明，新算法的搜索时间都不同程度地减少了，其中预测三步搜索算法和预测菱形搜索算法分别比三步搜索算法和菱形搜索算法的视频图像质量提高了；自适应菱形搜索算法恢复的视频图像质量也与全搜索算法恢复的结果非常相似。最后提出了一种基于DSP芯片的实时视频压缩系统方案，并进行了实现分析。

In visual IP telephony’s terminal systems, as the image for video information needs a mount of bits, high quality of channel bandwidth and large storage space are required, so video compression has been the effective ways which to solve this question. Motion estimate is the key skill in video compression coding, a way of between frames coding. It bases on the theory of temporal correlation between frames, to reduce temporal redundancy. An outstanding method of motion estimate is how to get an accurate motion vector. Finding out the most resembled part as the matching part in the neighbor by using the present part, pointing at it by motion vector, then finding out the different part between the matching part and the present part, at last rebuilding the compressed frame by the matching part and the different part. Therefore, in the video compression coding whose key skill is motion estimate, the speed of searching motion vectors and the accuracy of motion vectors are the most important factors to video effect. This thesis focuses on researching the classical algorithms of motion estimate, analying the advantages and disadvantages of exhaustive search algorithm, three-step search algorithm and diamond search algorithm, providing the scientific theory and experiment data. Basing on the relationship between the quality of video and the running time of algorithm and the motion vector of space-time correlation and center biased distribution, put forward predict three-step search algorithm, predict diamond search algorithm and adaptive diamond search algorithm, testify them on the experiment. The results indicate that the new algorithms are better than the former algorithms in searching time and compressing quality. Finally we design a scheme of real-time video compression system on the DSP chip and analyze them.