Feature Fusion Approach for Differently Exposed Images with Weighted Guided Filter

Abstract

Multi-exposure image fusion methodologies collect image information from multiple images and convey to a single image. Fusion with the aid of edge aware smoothing filters is a new treanding area. The difficultes of multi-scale processing and low level fusion operations are the main problems of the existing algoritms. In this paper we propose a novel multi-exposure image fusion method which uses a feature fusion method based on an edge aware weighted guided filter. Three important image features accounting for the quality of an image viz. contrast, sharpness and exposedness are extracted from the differently exposed input images and fused together to form a single saliency map which holds all the important information. A decision map is constructed for the fused feature and an efficient edge aware filtering technique called weighted guided filter is used for optimizing the obtained decision map. A two scale decomposition of input images is done in parallel with the initial feature extraction procedure. This decomposed image representation is fused with the optimized decision map to get the final result. The proposed method encompasses the advantages of simple two scale decomposition, optimization with edge weighting and simplicity of using a single fused feature. The experimental results and objective evaluations demonstrate that the proposed method can produce more accurate results with very good visual quality.