基于分布匹配的主动轮廓模型及其图像分割算法

针对运动目标跟踪的快速性和鲁棒性要求,提出了基于分布匹配的主动轮廓模型,该模型将模板和候选区域的分布匹配标准——相对熵(即Kullback-Leibler距离)作为主动轮廓的外部能量,使用目标和背景全图像的匹配,克服了跟踪过程中噪声的影响。同时根据匹配力构建帧内和帧间的运动向量,加快了序列图像分割速度。相比于现有的基于边缘、块匹配、区域匹配的主动轮廓模型,该模型具有更好的噪声鲁棒性及在跟踪过程中容忍目标旋转和较大位移的能力。实验验证了该模型的有效性。

Active contour model based on distribution matching and its image segmentation algorithm

A new active contour model based on distribution matching is proposed,In which the speediness and robust demands for motion object tracking are taken into account.In this model the Kullback-Leibler divergence is regarded as the external energy of the active contour.The influence of noise in the object tracking process can be overcome by using the distribution matching of object and background.Also the motion vector established from the matching force can accelerate the sequential image segmentation.Compared...