摘要
By combining multiscale stationary wavelet analysis with fuzzy c-means, a robust edge detection algorithm is presented. Based on the translation invafiance built in multiscale stationary wavelet transform, components in different transformed sub-images corresponding to a pixel are employed to form a feature vector of the pixel. All the feature vectors are classified with unsupervised fuzzy c-means to segment the image, and then the edge pixels are checked out by the Canny detector. A series of images contaminated with different intensive Gaussian noises are used to test the novel algorithm. Experiments show that fairly precise edges can be checked out robustly from those images with fairly intensive noise by the proposed algorithm.
借鉴平稳小波变换的多尺度分析思想,结合模糊聚类均值法,提出了一种高鲁棒性的图像边缘提取算法.该算法利用平稳小波变换的位移不变性,将小波分解后的分量进行配准构成一像素的特征向量,然后利用模糊c-均值进行无监督分类,分割图像,最后用Canny算子提取图像边缘.用一系列附加不同强度的高斯白噪声图像测试了该算法的有效性.实验证明在图像受到较强噪声(如附加高斯白噪声)污染时,该算法仍可检测到较好的边缘效果,展现出良好的鲁棒性.
