一种基于多尺度MRF模型的边缘检测算法

边缘检测中通常要处理两个问题：在单一尺度上检测和定位之间存在的矛盾；不同尺度上的边缘集成采用多尺度ＭＲＦ模型是有效的途径之一该模型引入线过程，用以表示边缘的确切位置，同时又能用任意适当的函数集成不同尺度上的边缘，本文根据ＧＢＦ不等式近似均值场，导出一组关于均值场的确定算式，地表示边缘间的相互作用以及模型参数对检测结果的影响，与文献〖１〗相比，需要较少的计算时间，并给出了用该算法处理几幅实际图像的实

An Edge Detection Algorithm Based on Multiscale Markov Random Field Models

Two problems usually need to be dealt properly in edge detection:the conflict between the detection and localization goals at one scale,and the integration of edges at all different scales.An efficient way is to use multiscale MRF models.Edge position can be precisely localized due to the introduction of a line process,and edges at all different scales can be integrated by using any proper function of the line process.In this paper,a deterministic system of equations is deduced by applying GBF inequality to approximate the mean field,which explicitly expresses the interaction between edges and the influence of model parameters on edge detection results,and can be solved with less computational time.Experimental results by applying the algorithm to a few real images are also given.