Image Fusion Based on Segmentation and Iterative Strategies

In this paper, a feature-level image fusion scheme is proposed that is based on minimal cut segmentation and an iterative strategy on the measurement of energy level of regions. Herein an energy model is defined and applied to evaluate the detail level of a region, and the more details in a region larger energy the region will have. Based on this idea, the fusion quality in a region segmented can be measured through the energy values. To achieve efficient fusion, an iterative procedure that selects different fusion methods is respectively used to perform fusion on each of regions of the image, and followed by an evaluation process until best fusion quality is evaluated. The proposed algorithm has been demonstrated by numerical experiment, and has shown many advantages like sensitive to error of registration and reduction of contrast over the pixel-based fusion algorithms.