Multimodal Medical Image Fusion Based on Hybrid L1- L0 Layer Decomposition Method

Abstract

Multimodal medical image fusion is the most important tool used for medical and clinical applications. The motivation of image fusion is to obtain more useful and detailed information into a composite image from the source image. The multiscale decomposition provides more flexibility and choice in selection of the appropriate fused images by varying from minimum to maximum level using selected rule. The most challenging task is estimating the wavelet transform decomposition levels which leads to the use of multiscale image fusion. The required operations are performed on fused image by varying the scale but selection of wavelet is the main issue. As the scale is high, fused image contains more details of source images. For preserving the more useful and appropriate information of poor contrast medical images, two scale image fusion is proposed. In this paper, hybrid $l_{1}-l_{0}$ decomposition based two scale fusion method is proposed for multimodal medical images. The image decomposition is done using hybrid $l_{1}-l_{0}$ decomposition model which produce base and detail layers. The detailed information of source images is identified using weighted average fusion rule. An average fusion rule is used for base layers which highlight edges, boundaries and contours. Image reconstruction is done using the linear combination of detail and base layers. The superiority of proposed method is compare with existing methods in terms of objective criteria like mean, standard deviation (STD), mutual information (MI), etc.