多尺度动静态联合彩色图像滤波处理

Multiscale dynamic and static filtering for color image processing

为在彩色图像降噪时有效保留图像边缘信息,给出一种动态和静态相结合的引导滤波算法。由静态引导滤波得到权重函数,利用动态引导滤波从正则化输入图像中得到邻域像素加权函数,将两者相乘,生成邻接矩阵;将邻接矩阵每列元素相加,置于对角线位置,构成度矩阵;用度矩阵与邻接矩阵做差,构造动态拉普拉斯特征矩阵,以此作为正则项,在平滑图像的同时保持边缘,实现图像多尺度动静态滤波。针对6幅彩色图像的测试结果表明,所给算法可有效获取不同尺度结构图像,且滤波后图像边缘较为光滑。

A dynamic and static guide filtering algorithm is proposed to preserve the edge information of the color image in the process of noise reduction. The weighting function is obtained by static guide filtering, the neighborhood weighted functions are obtained from the normalized input image by dynamic guide filter. The two kinds of functions are multiplied, and the adjacency matrix is obtained. The sum of the elements of each column of the adjacency matrix is placed in the diagonal position to form a degree matrix. The difference between the degree matrix and the adjacency matrix is used to construct the dynamic Laplasse feature matrix, which is used as the regular term. In this way, the multi-scale dynamic and static filtering can be realized to smooth the image and keep the edge. The test results of 6 color images show that, the proposed algorithm can obtain the image with different scale structure, and the edge of the filtered image is smooth.