基于小波变换的多尺度图像边缘检测

Image Edge Detection Based On Multiscale Wavelet Transform

边缘作为图像的最主要特征,成为图像信息获取的重要内容.而小波变换具有检测局域突变的能力,而且可以结合多尺度信息进行检测,因此成为图像信息边缘检测的优良工具.文章首先构造了高斯多尺度边界检测算子,然后根据信号边界与噪声边界的小波变换模值跨尺度传递的不同特性,讨论了不同尺度的检测算子检测的边缘所具有的特点,在此基础上提出由边缘传递、继承和生长构成的多尺度边缘关联融合算法.实验结果说明这种特征提取方法不仅有效地降低了噪声,而且融合的边界比较完整,定位准确.

The edge, as the most basic characteristic of images, is an important content of obtaining information of a picture. The wavelet transform can detect part mutation , and can do it combining multiscale information, so wavelet has become a good tool of detecting information of edge images. In this paper,a multiscale Gaussian edge detector is constructed. According to transfer properties across scales of the wavelet modules of the signal edge and the noise edge, we combine the propertics of edges in different scales and propose a muhi-scale edge fusion algorithm consisting of edge transfer, edge inherit and edge growth. The resuh of experiments shows that this algorithm can get rid of the affect of noise and the edges fused have precise position and intact contour.