多尺度边缘检测算子的构造

图像处理中,边缘检测具有相当重要的作用,它可作为模式识别、图像分割及图像场景分析的基础。通常一个性能良好的边缘检测算子应该是具有以下两个特征的滤波器:首先,必须是一个微分算子,即一阶或者二阶微分;其次,算子大小能够随意缩放,以便于用较大的算子检测边缘,而用较小的算子检测细节特征。根据这一思想,构造了多尺度的边缘检测算子,并进行了实验,得到的图像边缘具有相当好的效果。

Conformation of Multiscales Edge Detection Operator

Image edge detection plays a very important role in image processing. It can be used as the base on which modal distinction, image division and image scene analysis are made. A favorable edge detection operator is a filter that should have two character usually. First, it must be a differential coefficient operator, that is one or two rank differential coefficient. Secondly, the operator is able to random zoom, in order to use better operator detect edge, and use lesser operator detect detail character. The paper conforms multiscale edge detection operator based on the idea of frontal saying, and makes some experiment, and the image edge by experiment has excellent effect.