Feature-Level Image Fusion Using DWT, SWT, and DT-CWT

Abstract

Image fusion is the process of combining information from two or more sensed or acquired images into a single composite image that is more informative and becomes more suitable for visual processing or computer processing. Image fusion fully utilizes much complementary and redundant information of the original images. The aim of image fusion is to integrate complementary and redundant information from multiple images to create a composite image that contains a better description of the scene than any of the individual source images. Feature-level image fusion (FLIF) algorithms (both in spatial and in frequency domain) were developed and evaluated using fusion quality evaluation metrics. The images to be fused are passed through joint segmentation algorithm to get the common segmentation map. Salient feature, viz. standard deviation, is computed for corresponding segments (both the images), and the segment was chosen based on best salient feature. It was done for all the segments. Five different image sets were used to evaluate the proposed fusion algorithm. To compare the performance of this algorithm, three different pixel-level image fusion algorithms, viz. DWT, SWT, and DT-CWT, were also implemented and evaluated. From this study, it is concluded that FLIF provides a good fused image at the cost of execution time and also it requires a good segmentation map. Most of the time DT-CWT provides good fusion results since it considers the edge information in six directions. In all cases, the DWT-based pixel-level image fusion algorithm does not provide good results since it does not consider the edge information and lack of shift invariant. SWT-based image fusion algorithm provides good results in some cases where there are no much edges in the images to be fused, it is shift invariant, and it does not consider the directional edge information.