利用后验均值提取电视电话图象边缘的方法

An Algorithm of Edge Detecting Using Mean Posterior for Video Telephone Images

本文以马尔可夫随机场模型研究了电视电话图象．利用后验均值建立了提取人物头肩象轮廓（包括内轮廓和外轮廓）的方法．该方法针对此类图象的特点定义了合适的边素，给出了相应的邻域系、连通子图及能量函数．此外还根据人眼辨别灰度极的特性在叠代前对原始图象进行了均衡．由于采取了这些措施，使得该算法不仅提高了轮廓的定位精度。

In this paper, images of video telephone are investigated by the Markov Random Field (MRF) model. An algorithm for detecting the silhouette (including inner silhouette and outside silhouette) of head shoulder image is established with mean posterior. Allowing for the characters of this type of images, proper edge element and its neighborhood system, subset and power function are defined. Original image is also balanced based on the human eye features of recognizing gray scalars. With these methods, not only the precision of silhouettes location, but also the ability of noise resisting are improved.