Multi-Exposure Image Fusion Using Edge-Aware Network

The natural dynamic range is much broader than the dynamic range a digital camera can capture. As a result, the images captured by ordinary cameras suffer from visibility issues; some images look overly bright, while others look overly dark. Therefore, in the research article, we propose a solution for image fusion, where multiple images can be combined to produce a single comprehensive image with better perceptibly. To achieve that, we designed a feature extraction block, which can extract features of an image at different levels and combines them to get more refined and detailed feature maps. In addition to that, we also designed a custom layer to extract edges, and those edges later are combined with the feature maps to give a fused image. We evaluated and compared our proposed architecture with some of the latest architecture, designed for the same reason, in terms of Image Structure Similarity and Human Perception-Based metrics. Experimental results corroborate that the proposed scheme outperforms the contemporary techniques for image fusion.