关键词: 原木端面, 水平集, CV模型, 局部图像拟合（LIF）模型, 圆形约束

Abstract: The paper, in allusion to the segmentation problems of log end image obtained in natural conditions, and according to the characteristics of the log end image, improves the traditional CV（Chan and Vese） model, and fits the interior of evolving contour with gradient information. At the same time, with reference to the Local Image Fitting（LIF） model, and joining with circle prior knowledge, it proposes an improved active contour model based on circle dependent, which is called CV-LIF. The model combines global and local energy together to restrain the evolution of contours. On the basis of pre-segmentation image, it represents the regions which are to be segmented with multi-level set, and does re-segmentation to each level set region with the model proposed. This model solves the problem of inaccuracy of multiple logs ends image segmentation under the complex background. Experiments for single log end and multiple logs ends segmentation indicate that the model can segment the logs ends correctly, and has better anti-noise performance. It can effectively overcome the interference of background, and runs faster.

Key words: log end, level set, Chan and Vese（CV） model; , Local Image Fitting（LIF） model, circle dependent