块匹配算法中特征块的选取方法研究  

Research on Choosing Feature Block in Block Matching Algorithm

作　　者：江小明[1] 杨德贵[1] 魏玺章[1] 

机构地区：[1]国防科技大学电子科学与工程学院空间电子与信息技术研究所,湖南长沙410073

出　　处：《现代电子技术》2009年第4期117-120,123,共5页Modern Electronics Technique

摘　　要：运动背景补偿是红外序列图像目标跟踪的基础,首先分析块匹配法原理及其误匹配现象产生的原因;然后提出3种定义块的复杂度的方法,据此快速选取用于背景运动估计的特征块群;最后经过外场实测图像验证,基于图像梯度特征定义块的复杂度能够比较准确地估计出背景的运动矢量,且性能优于其他两种方法。The motion background compensation is the foundation in the infrared sequence image target tracking. Firstly, the block matching principle and the reason of mismatching are introduced, Then, three kinds of defining methods to the complexity of block are presented, so feature block group applied the background motion estimation is chosen quickly. Finally, verifying by the sociable reality data, what the complexity of block definition based on the image gradient characteristic can be fairly more accurate estimates a motion vector of background,and the function is superior to others.

关 键 词：背景运动估计 块匹配 图像梯度 块的复杂度 特征块群 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]