基于小波域内层间系数反向相关提取图象边缘

Wavelet Using Interscale Back Correlation for Edge Detection in Noisy Image

为了更好提取出边缘,提出利用小波层间系数之间的相关性通过反向移位构建滤波器从而提取图像边缘.首先对含噪图像进行小波分解,将相邻各层小波系数进行反向移位相关,阈值化后作为滤波器对平滑小波系数有选择地进行滤波操作,去除噪声影响,得到图像的边缘图.利用此方法对256×256×8bit/像素的Rice图像做了边缘提取,边缘提取准确,具有更好的降噪性能同时也降低了计算量.

In order to obtain edge information, an edge detection method in noisy image using back shifting wavelet coefficients of correlative scales to construct a filter and produce the edge image is presented. First,the noisy image is decomposed into different scales wavelet coefficients including horizontal, vertical and diagonal direction,then, the shifting wavelet coefficients are back multiplied step by step . After thresholing,a filter constructed by the coefficients will be selectively applied to an original level in every direction to depress noise and produce the edge image. Experimental results prove that this approach not can detect edge information correctly,get rid of noise very well and depress the number of computation