A New Polarized Image Fusion Algorithm Based on Two-scale Guided Filtering

This paper proposes a new fusion algorithm based on two-scale guided filtering for polarized images. After generating the detail and base layers from source image, we construct the saliency map. It provides good characterization of the saliency level information, which is beneficial to the fusion. Subsequently, two different scales of guided filtering are performed on the weight map which is derived from the saliency map. Finally, weighted fusion method is used with the filtered weight map to fuse the detail layer and the base layer. In the proposed algorithm, we improve the saliency map by combining second-order and first-order difference edge detection operations. The weight maps constructed with the improved saliency map can fulfill the requirements of the new weighted fusion method. Compared with the original method, all of the benchmarks on testing images have been significantly improved during objective evaluation. Subjectively, it is also shown that the details of the improved saliency map have become obviously sharper.