Description

Here we provide a novel approach in the fusion based on similar region. For the image fusion we implement the watershed segmentation algorithm. In initial step we find the structure similarity of two images. SSIM contribute to its quality evaluation. The luminance comparison respects the pixel intensity values. Similarly, the contrast and the structure comparisons clearly assess the edges between evaluated image and reference. Therefore, they can reflect the correlation of two images in terms of intensity, edges, and structure. The luminance map (l-map), contrast map(c-map), structure map(s-map), and SSIM index map (SSIM-map) for two source images are collectively referred to as similarity maps. Then we generate the gradient of the SSIM map. Then we get the over segmented result by watershed segmentation. After over segmentation we produce the fine segmentation result based on similar region. Experimental results confirm that our region-based fusion method achieve superior results.