SURF中快速求取积分图像的直接替换算法

Direct Replacement Algorithms of Fast Computing Integral Image in SURF

SURF通常采用的积分图像求取算法存在内存消耗大、时间花费多的问题。文中以降低内存消耗和时间花费为目的,分别对两种求取积分图像的Viola算法和Lewis算法进行改进,并结合SURF算法只基于积分图像计算而不使用原图像的特点,提出了两种只占用一个图像空间的直接替换算法。经实验分析与结果验证,这两种直接替换算法在执行时不仅大大降低了内存消耗,而且其时间花费均明显少于Viola和Lewis算法。

Usually the integral image computing algorithm used by SURF has problems of costing more memory and time. In order to solve this problem, two integral image computing algorithms Viola and Lewis were improved in this paper. Combining the characteristic of SURF algorithm that using integral image rather than original image to compute, the two direct replacement algorithms which just cost one image space were proposed. Verified by experimental analysis and results, these two direct replacement algorithms not only reduce memory when computing, but also cost less time than Viola and Lewis algorithms.