利用小波多尺度积实现裂纹缺陷的边缘检测

Edge detection of crack defect based on wavelet multi-scale multiplication

提出了一种基于小波变换的图像边缘检测方法,即利用边缘信息的多尺度特性和小波变换模极大值对图像进行多尺度分解,将相邻尺度的小波系数相乘增强边缘,再通过双阈值去噪的方法,得到最终的图像边缘。实验结果表明该方法很好地解决了噪声和坏边的问题,边缘连续的同时又保证了边缘定位的准确性,采用双阈值的算法明显优于采用单阈值,可以有效用于结构件的检测。

A new image edge detection algorithm which is based on wavelet transform is concerned with doing multi-scale wavelet decomposition to the image with multi-scale of edge information and the modulus maximum value of wavelet transform,and then making a multiplication of consecutive scale wavelet coefficient to enhance edge and achieving the final image edge with double threshold to remove noise components.The result of test indicates that this algorithm solves problems of edge noise and the bad edge, and ensures the accuracy of edge continuation and positioning.The algorithm with double threshold is superior to a single threshold and can be used effectively in framing member detection.