摘要
本文提出了一种Bilateral滤波器的快速算法。在算法的设计上,首先对空间邻近度函数建立一组动态掩模,代替原始耗时的逐点运算,时间降至原有时间的三分之二;再离散化亮度相似度函数,使整个Bilateral滤波器计算公式形成卷积形式;在算法的实现上,引入了快速傅立叶变换FFT进行加速,时间复杂度由卷积运算的O(N2)降至O(Nlog2N),降低了一个数量级。
This paper proposed a fast approach for Bilateral filter, First, instead of the time-consuming point-by-point calculation in the primal, we set up a set of dynamic templets to the geometric closeness function,as a result, only cost two-thirds computing time than before . Then, We discretize the photometric similarity function to make the computational formula of the Bilateral filter come into being a convolution. In the end, since the discrete Fast Fourier Transform (FFT) have cost O (Nlog2N), the O (N2) convolution can be greatly accelerated using discrete FFT,there is a gain of one order of magnitude.
出处
《微计算机信息》
北大核心
2008年第27期213-215,共3页
Control & Automation