一种基于t图的多尺度边缘检测方法

在多尺度边缘检测方法中，滤波器的滤波尺度的选取是关键，本文通过检验一给定像素邻域的灰度分布来判断此像素是否位于平滑区，并产生一幅五维向量图－t图，然后根据t图自适应地选取小波变换的尺度。实验结果表明，应用这种方法进行多尺度小波边缘检测具有较强的抗噪能力。

Mutiscale Edge Detection Based on t-diagram

The most important thing in multiscale edge detection is the filter scale. By applying the Student t-test to compare the distribution of the intensities in the neighborhood of a given pixel, the pixel is classified to be either in an edge region or in a homogenous region. Accordingly, a five-dimensional vector diagram (t-diagram) is obtained. The Harr wavelet whose scale is adjusted by t-diagram is used to detect the edge. Experimental results show that using the proposed method, edges of natural images can be robustly detected despite of presence of additive noise.