海量微小差异图像群中智能边缘检测方法仿真

摘要

An image edge detection method based on Canny arithmetic operator,non-downsampling Contourlet transform and fuzzy C mean clustering method is proposed.The multiplicative noise of massive small difference images is transformed into additive noise.Combined with Canny operator,the edge direction in difference image group is calculated.According to characteristics of rich edge and smoothing region in massive small difference image,and the diversity of difference image in gray level,the division of gray difference is completed by combining with the regional growth.The image in image group is decomposed into low-frequency components and high-frequency components through non-downsampling Contourlet transform,and the edge information is extracted.The fuzzy C-means clustering method is used to cluster edges,so as to obtain the edge image with low frequency.For each subband of high frequency component of small difference image with much edge detail information,the edge is detected according to the modular maximum,and the edge of image is reduced,then details of small difference image are enriched.Through integrating low-frequency components and high-frequency components in massive small difference image,a complete image edge is obtained.Simulation results show that this method has a good ability to suppress Gaussian noise and has good positioning accuracy of image edge extraction precision and interference robustness.%由于海量图像群中会同时含有物体边缘、阴影以及噪声问题,采用常规图像边缘检测方法很难从噪声或是微小差异特征中区分出精确边缘,检测出的边缘存在细节损失严重、边缘不连续以及噪声去除不完全等问题.提出一种结合Canny算子、非下采样Contourlet变换以及模糊C均值聚类方法的图像边缘检测方法.将海量微小差异图像乘性噪声转换为加性噪声形式,结合Canny算子计算差异图像群中边缘方向.根据海量微小差异图像边缘丰富、区域平滑的特点,以及差异图像在灰度上的差异,结合区域生长完成灰度差异的分割.通过非下采样Contourlet变换将图像群中图像分解为低频分量和高频分量,提取边缘信息,采用模糊C均值聚类方法对边缘进行聚类获得低频边缘图像,对于边缘细节信息较多的微小差异图像高频分量各个子带,依据模极大值检测边缘,减少图像为边缘,丰富微小差异图像细节,通过对海量微小差异图像群中低频分量和高频分量进行融合获得完整的图像边缘.实验结果表明,上述方法对赢斯噪声具有较好的抑制能力,具有良好的图像边缘定位精度和干扰鲁棒性.