Abstract: The effects of accuracy and efficiency play an important part in the process of image stabilization.By analyzing image stabilization principles and commonly used algorithms，a new algorithm which combined gray-scale projection with block-matching algorithm is proposed.The algorithm usess gray-scale projection algorithm to perform the first matching and rough compensation，in which a new scheme of searching called Fast Search（FS） is proposed to decrease computation.Here，gray-scale projection plays a role of making assistance for the next steps.Then it uses block-matching algorithm for the second matching and precise compensation，which Diamond Search（DS） algorithm is employed for improving efficiency greatly.Therefore the algorithm is a coarse-to-fine process for image stabilization.The combination of gray-scale projection and block matching algorithm can work very well.Experimental results show that the new algorithm can preserve high stabilization accuracy，at a dramatic reduction of computational complexity.

Key words: video image stabilization, gray-scale projection, block matching, Fast Search（FS）, Diamond Search（DS）

摘要： 视频稳像的精度和处理效率在视频稳像处理中非常关键。通过分析视频稳像的原理和常用的几种算法，提出一种基于灰度投影和块匹配相结合的算法。先使用灰度投影算法进行首次匹配工作和粗补偿，针对灰度投影提出一种快速搜索算法，可以有效减少计算量。灰度投影算法的作用是为下一步的工作做铺垫。然后使用块匹配进行第二次匹配和精确补偿，其中菱形搜索（DS）策略的应用可以有效地提高算法效率。提出的算法体现了一种由粗到精的处理思路，灰度投影和块匹配算法可以很好地配合工作。实验结果表明，该算法在保持较高稳像精度的同时有效降低了计算复杂度。

关键词: 视频稳像, 灰度投影, 块匹配, 快速搜索, 菱形搜索