基于形变模型的图像分割和非刚性目标的跟踪  

IMAGE SEGMENTATION AND TRACKING NON - RIGID TARGETS BASED ON DEFORMABLE MODELS

作　　者：夏利民[1] 谷士文[1] 沈新权 

机构地区：[1]长沙铁道学院电子系,长沙410075 [2]Department of Electrical Engineering,University of Surrey,Guildford UK

出　　处：《模式识别与人工智能》2000年第1期64-68,共5页Pattern Recognition and Artificial Intelligence

基　　金：湖南省自然科学基金

摘　　要：本文在活动轮廓的基础上,提出了基于形变模型的图像分割和非刚性目标的跟踪方法。为了克服活动轮廓的一些缺陷,引入了新的与图像特性有关的外部能量和松弛法;将形变模型和光流模型相结合用于非刚性目标跟踪。该方法对噪声不敏感,能准确、可靠地检测物体曲率高的边缘部分,跟踪非刚性运动目标。实验结果验证了该方法的有效性。Image Segmentation and tracking non-rigid targets based on deformable models is presented in this paper. In order to overcome some shortcoming of active contours, we introduce a new external energy and a relaxation. The approach uses a combination of deformable models with image optical flow for tracking non- rigid targets. The proposed method is able to locate high curvature parts of object accurately and tracking non- rigid targets. Experimental results are also included to show the feasibility of our newly presented method.

关 键 词：形变模型 边缘检测 图像分割 非刚性物体跟踪 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]