摘要： 工程应用中需要处理的信号通常都局限在有限的时域或频域之中。在处理这些数据时，多数变换方法都要求在边界处做某种延拓处理，小波变换也不例外。而传统的延拓方法在前后两个边界处都会产生伪边缘，并且在延拓时还需考虑滤波器系数长度的奇偶性。为了解决这些问题，文章对图像小波边缘检测中传统的边界处理方法进行了研究，并对其进行了改进，即采用右边界延拓法消除了左边界处的伪边缘，减少了伪边缘的数目。在此基础上又进一步提出了右边界右端点延拓法，在边缘检测中取得了非常好的效果，不仅消除了传统边界处理方法产生的伪边缘，而且在延拓时不需要考虑滤波器系数长度的奇偶性，为实际工程应用带来了方便。

关键词: 小波变换, 边缘检测, 边界处理

Abstract: The signals from project applications are usually localized in the finite time and frequency domains, so most transform methods need to extend their boundaries. But traditional border processing methods will cause false edges, and also should consider the parity of filter coefficient. To solve these problems, this paper analyzes and improves the traditional border processing methods on wavelet image edge detection, and also gives a new kind of border processing method on the basis of the improved method, it gains an superduper effect on wavelet image edge detection. There are no false edges when the new kind of border processing method is used, further more, it doesn’t need to consider the parity of filter coefficient, all of these bring conveniences for the practical work.

Key words: Wavelet transform, Edge detection, Border processing