Content area
Full Text
Abstract: For pixel-level image fusion, the edges of the source images should be as much as possible integrated into the fused image because the human visual system is sensitive to them. In this study, the authors utilise the multiscale edge-preserving decomposition (MSEPD) based on the weighted least squares filter to fuse the source images. In the authors' method, first, the source images are decomposed by the MSEPD into a base image and a series of detail images, respectively. Then, the detail images of same scale are combined via the different fusion rules designed for different kinds of source images; the base images are combined via the average-value rule. Finally, the fused image is constructed by adding the fused base image and detail images together. The proposed fusion method is verified on several kinds of images and compared with some methods based on multiscale decomposition. The experimental results indicate that the proposed method can provide better fused images, meanwhile manifesting a good edge-preserving performance.
(ProQuest: ... denotes formulae omitted.)
1 Introduction
The development of the imaging technique enables a single system to use multiple imaging devices and accordingly acquire multiple images of the same scene, which are always mutually complementary and thus make the system more robust. However, this kind of system may have a limited efficiency because these images need to be processed individually. For improving its efficiency, the system often employs the technique of image fusion to merge multiple source images into one [1]. The actual image fusion can be accomplished at different levels: pixel-level, feature-level and decision-level [2], at which the information representations of an image have different meanings. Pixel-level image fusion is able to retain more original information [3], compared with feature- and decision-level. The results of the pixel-level fusion need to satisfy some generic requirements [4]: (i) all visual information contained in the source images should be as closely as possible preserved in the fused result; (ii) any irrelevant features such as artefacts should not be introduced into the fused image and (iii) the noises in the source images should be maximally suppressed in the fused image. In addition, the edges of the images can effectively stimulate the visual perception of humans, and thus are vital for interpreting the...