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，而且具有很高的抗噪声能力 In this paper, we study the videophone image based on Markov random field model. A posterior mean value is used to establish the method of extracting head and shoulders contour (including inner contour and outer contour). The method defines suitable edge factors for the characteristics of such images, and gives the corresponding neighborhood series, connected subgraphs and energy functions. In addition, according to the characteristics of the human eye to distinguish grayscale of the original image before the iteration equalization. As a result of these measures, the algorithm not only improves the positioning accuracy of the contour, but also has a high anti-noise ability