中文摘要：

提出了一种基于MAS小波变换多尺度相关的噪声污染图像边缘检测方法.该方法采用二进MAS小波对图像进行多尺度分析,由于小波变换的模极大值充分刻画了图像的奇异点,利用模极大值得到所有的奇异点包括边缘和噪声的图像;利用边缘和噪声具有不同的Lipschitz指数造成它们的小波变换模在不同尺度下的不同传播特性,根据小波变换模尺度相关性区分边缘和噪声,得到边缘轮廓不太光滑的图像;将两幅图像进行融合,得到最终检测结果.实验结果表明,该方法能够有效地对噪声污染的图像进行边缘检测.

英文摘要：

Based on the modulus angle separated（MAS）wavelet transform,a new algorithm of edge detection for image corrupted by noise has been presented.The discrete dyadic MAS wavelet transform is employed to produce the multi-scale representation of image and multi-resolution analysis.The wavelet transform modular maxima are representations of singular points,which include edges and noise.Since edges and noise have different characterizations of Lipschitz exponents,they have different representations in different scales.Normalization correlations moduli are calculated and compared with moduli of wavelet transform to separate the edges and the noise.Then the modular maxima map of small scale and the coarse edge map are combined to obtain the real edge.Experiments turn out the algorithm works efficiently in edge detection for image corrupted by noise.