摘要
研究图像融合的问题。针对传统基于像素级别的图像融合算法特征单一,或者基于区域性质的融合算法的缺陷,提出在小波变换域进行滤波,然后分割成若干区域,在不同区域进行图像融合的算法。首先,将图像变换到小波频率域;然后分别利用低频和高频子图像,通过boost滤波方法获得一个滤波图像;在滤波后图像上使用基于图论的分割算法获得不同的图像区域,最后,根据本文定义的两个融合准则进行区域内融合处理获得最终的融合图像。所提方法只需进行一级小波变换,较其他基于小波变换的图像融合算法具有更高的运算效率,同时,利用boost滤波技术,在保持图像细节的同时,能最大程度地保证各个不同源图像中相同信息得以保留。所提方法不容易受到噪声的干扰,融合结果更柔和,能广泛使用在图像预处理系统中。
Image fusion approach is studied in the paper. Traditional pixel level-based image fusion algorithm is of single in characteristic, in light of this, or of the defect of region property-based fusion algorithm, we propose an algorithm, in it the filtering is done in wavelet transform domain followed by segmenting into a couple of regions, and the image is fused in different regions. First, the image is transformed into wavelet frequency domain; Secondly, the boost filtering method is applied on sub-images with low and high frequencies respectively to obtain a filtered image; Thirdly, on the filtered image the graph theory-based segmentation algorithm is utilised to gain different image regions; At last, according to two fusion criteria defined in the paper, final fused image is derived from fusion processing within the region. The proposed method just makes one-order wavelet transform and so has higher operation efficiency than other wavelet transform-based image fusion algorithms. Meanwhile, by utilising boost filtering technology, the same information in images with different sources can be preserved to most guaranteed extent while keeping the image detail. The method is insusceptible to the noise interference and has gentler fuse result, it is able to be widely applied in image pre-processing systems.
出处
《计算机应用与软件》
CSCD
北大核心
2013年第7期290-293,共4页
Computer Applications and Software
