Unsupervised Real-Time Mobile-Based Highly Dark Image Texture Enhancement App

Detection of cracks or defects in inaccessible areas of a machine, analysis of disaster-stricken regions via aerial photography, and unmanned aerial vehicle-based rescue operations in hazy or smoky conditions all require the use of modern cameras. Enhancement algorithms that can extract the hidden information in under and over-exposed regions of an image are the need of the hour. The proposed method is the first unsupervised method in the literature that enhances more than one type of degradation, viz., global and local contrast, under- or over-exposed images, and extremely low-light images. The paper utilizes local binary patterns to extract concealed textures in these regions. A novel probabilistic transfer function improves the perceptual quality of these regions. The cooccurrence matrix is used to make the algorithm computationally efficient in order to deal with the high memory consumption of modern devices’ images. Exposure fusion is used to fuse the texture-enhanced and color-vibrant images. The enhanced image provides superior perception and the best average value of quantitative parameters as compared with state-of-the-art methods. An industry-focused, platform-independent app is created and made available to the general public for additional study.