一种大尺度Gauss模糊的快速采样算法

针对大模板Gauss模糊的传统算法计算复杂度较高的缺点,提出了一种基于快速采样的新算法,将计算复杂度从O(N2)降到O(N)级别.利用极大似然估计理论,对传统Gauss模板和采样模板进行分析,给出了采样算法和传统算法的联系与区别.实验表明,新算法在极大的提高算法的速度的同时,可以将两者之间的能量误差控制在1%左右,当采样点等于10N时,人眼基本无法察觉两种算法的差异.

A fast Gaussian blur with large scale mask based on sampling theory

High computational complexity is a great disadvantage for traditional algorithm of Gaussian blur,especially where large scale mask involves.A new algorithm is proposed based on fast sampling theory,which reduces computational complexity from O(N 2) level to O(N) level.A thorough analysis is applied between traditional mask and sampling mask by using the max likelihood estimation theory.The analysis results reveal the connections and differences between the new algorithm and the old one.Experiments show that...