基于图像融合的多分辨率动态轮廓线检测跟踪  

Detection and tracking scheme of multi-resolution dynamic contour based on image fusion

作　　者：赵鹏[1] 浦昭邦[1] 张田文[2] 

机构地区：[1]哈尔滨工业大学自动化测试与控制系 [2]哈尔滨工业大学计算机科学与工程系,哈尔滨150001

出　　处：《哈尔滨工业大学学报》2007年第11期1690-1694,共5页Journal of Harbin Institute of Technology

基　　金：国家自然科学基金资助项目(60075010)

摘　　要：提出一种基于图像融合的动态轮廓线检测与跟踪方法.该方法首先使用一种基于对比度塔形分解的图像融合技术.其次,对于融合图像的各分解层,使用一种新颖的多分辨率动态轮廓线.这种方法有3个优点,首先,它不但继承了单分辨率动态轮廓线的优点,而且降低了对初始轮廓线的要求.此外,推导了不同分辨率情况下动态轮廓线与其允许的目标运动速度的关系公式,并用这些公式定量计算出单分辨率和多分辨率情况下动态轮廓线允许目标运动的最大速度,证明本方法可以跟踪运动速度较快的目标.最后,它充分利用了融合图像具有的更准确的图像信息,提高了动态轮廓线检测与跟踪的准确性.A detection and tracking scheme of a dynamic contour based on image fusion is proposed. Firstly, an image fusion based on contrast pyramid decomposition is applied in this scheme. Secondly, for every hierarchical fused image, a novel multi-resolution dynamic contour is applied. This scheme has three merits. Firstly, it not only retains the advantage of the mono-resolution dynamic contour, but also cuts down the strict requirement to an initial contour. Secondly, the relation formulas for a dynamic contour and the allowed tar- get＇ s velocity in different resolutions are derived. According to these formulas, the allowed maximum target＇ s velocity is calculated quantitatively under mono-resolution and multi-resolution cases so that it proves that this scheme can track a target with faster Speed. Finally, the more accurate image information of the fused image is applied in this scheme, which increases the detection and tracking accuracy of a dynamic contour.

关 键 词：图像融合 动态轮廓线 多分辨率 对比度塔形分解 

分 类 号：TN911.73[电子电信—通信与信息系统]