基于带权重像素平移的图像滤波快速实现方法

Fast implementation of image filtering based on weighted pixel translation

针对图像滤波中存在冗余计算的问题,提出一种基于带权重像素平移的图像滤波快速实现方法.首先,将滤波器权重与像素的乘积作为带权重像素,根据权重在滤波器中出现的不同位置,将带权重像素累加至对应的滤波器中心所在图像像素;其次,通过减少线性滤波器中不同位置由相同权重带来的冗余乘法运算,减小计算量,提高计算速度.实验结果表明,所提方法实现的高斯滤波比直接实现法计算速度快6~7倍,可有效实现图像的快速滤波.

A fast implementation of image filtering based on weighted pixel translation is proposed to solve the problem of redundant computation in image filtering.The method firstly takes the product of filter weight and pixel as the weighted pixel,then adds the weighted pixel to the corresponding filter center according to the different positions of the weight in the filter.This method improves the computation speed by reducing the redundant multiplication caused by the same weight in different positions of the linear filter.The experimental results show that the Gauss filter realized by this method is 6~7 times faster than that realized directly,which shows the effectiveness of the fast image filtering method proposed in this paper.