Multi-exposure image fusion

Abstract

The high dynamic range (HDR) present in nature severely affects digital photography and produces saturated regions in captured images. This thesis proposes a novel pixel-wise multi-exposure image fusion algorithm to overcome the problem of HDR in digital photography. The proposed algorithm skips the tone mapping process by computing the low dynamic range (LDR) fused image that can be directly displayed on standard dynamic range (SDR) devices. Mathematical morphological profiles are used to extract visible structures from LDR image sequence at multiple levels, through which optimal weights are computed for effective fusion process. These weights are the deciding parameter for the blending process, multi-scale weighted summation method is used as blending technique. The most important finding of the proposed technique is the computation of weights through extracted structures at multiple levels, which results optimal weight maps for fusion process and ensures maximum structure preservation. A large number of data set, including natural (indoor and outdoor, still and human life) scenes are used to evaluate the performance of the proposed method with diverse existing techniques. Experimental results demonstrate the superiority of the proposed algorithm over many advance existing MEF techniques. Keywords: High Dynamic Range Imaging, Multi-Exposure Image Fusion, Multi- Scale Mathematical Morphology, Difference Profiles.