作　　者： ; ; ; ;

机构地区： 宁波大学信息科学与工程学院

出　　处： 《光子学报》 2007年第8期1543-1547,共5页

摘　　要： 针对多视点图像系统中视点图像颜色不一致的问题,提出了基于区域分割与跟踪的多视点视频校正算法.采用3维高斯混合模型来描述颜色信息,利用期望-最大算法来精炼模型参量,并通过自适应聚类和概率平滑操作来得到精确的分割图像.这样建立起多视点图像分割区域间的局部映射关系,并利用这种局部映射关系对全局图像进行校正.最后利用视频跟踪技术实现对视频图像的校正.实验结果表明,该算法能消除颜色过饱和或颜色区域混淆的影响,且具有较好的校正效果. In order to solve non-consistence color appearance correction algorithm based on region classification and tracking adequtely characterized by a Gaussian Mixture Model and between different viewpoints, a multi-view video is proposed. Color statistics in each region can be Expectation-Maximum is used to refine model parameters. Then accurate segmentation image can be obtained by adaptive clustering and probability smoothness operation. Global correction can be achieved by the local mapping relation between the best similar regions in reference image and current viewpoint image. Finally video tracking technique is performed to correct multi-view video. The experimental results show that the proposed algorithm can effectively eliminate the influence of color over-saturation or color region mixture, and achieve good correction result.

关 键 词： 图像校正 区域分割 多视点视频 期望 最大

领　　域： [电子电信] [电子电信]