Perceptual Quality Evaluation for Exposure Fusion Image

In this paper, a new approach for automated perceptual quality evaluation for pyramid-based exposure fusion is presented. A machine learning method, using neural network training is used. The obtained results are verified using both established objective measures, and subjective evaluation of the same set, acquired by mean opinion score survey on multiple human subjects. The main advantage of the proposed approach is that it can grade the perceptual quality of the fused image even prior to the completing of the fusing process, in the stage of pyramid decomposition. This also allows implementation of a mechanism that could stop the pyramid decomposition in the most optimal pyramid level, in order to achieve the highest quality output image.