摘要
针对传统的基于像素点和窗口策略的融合算法对图像特征表征的失真,提出了一种基于区域分割的序列图像融合算法。首先将序列红外图像分割为3个不同的特征区域,目标区域、背景区域以及灰度区域,并将分割结果映射到可见光图像中。随后,利用多尺度几何分析工具非下采样Contourlet变换(NSCT)有效提取图像特征的特点,根据不同区域的特性在NSCT域设计不同的融合规则。对试验结果进行主观和客观的对比,结果表明:基于区域分割的序列图像融合算法不仅能够为融合图像保留更全面、丰富的背景信息,还能够更加有效、准确地提取图像中的目标特征。该算法优于传统的基于像素点和窗口规则的融合算法,是一种有效可行的图像融合算法。
Aimed at the drawback of traditional fusion methods based on pixel and window strategy, which have not the ability to express the characters of fused image efficiently, a fusion algorithm of sequence infrared image using region segmentation was proposed. Firstly, the sequence images were divided into target area, background area and gray area. Then these different areas were mapped into the visible images. According to the characters of the different areas, the different rules were designed in nonsubsampled Contourlet transform （NSCT） domain. The NSCT could provide a flexible multiresolution, local and directional image expansion, and a sparse representation for 2-D piecewise smooth signals, and then different fusion rules were applied to fuse the NSCT coefficients for given regions and optimize the quality of the fused image. Experimental results were compared both in subjective and objective standards. It is showed that the fusion algorithm not only keeps the background information of fusion image completely and richly, but also extracts the target characters of image accurately and effectively. The proposed algorithm is superior to conventional fusion methods, and is feasible and effective.
出处
《红外与激光工程》
EI
CSCD
北大核心
2009年第3期553-558,共6页
Infrared and Laser Engineering
基金
国家自然科学基金资助项目(60802084)
作者简介
刘坤（1982-），女，河北唐山人，博士生，主要从事图像融合等方面的研究。Email：cc_liukun@163．com
郭雷（1956-），男，山东海阳人，教授，博士生导师，主要从事神经计算、视觉计算、图像和视频处理以及模式识别等方面的研究。Email：lguo@nwpu．edu．cn
