A Comprehensive Review of Convolutional Neural Network based Image Enhancement Techniques

Image enhancement is the most important field in image research and processing. The main aim here is to artificially enhance the photo quality and visual appearance. Many photos end up with unusually high amount of noise and grain due to poor sensor quality and control, or with low sharpness. It is expensive and sometimes impossible to increase the resolution of the sensors (even counter intuitive at times) inside the cameras. Convolutional Neural Networks (CNN) rely on image structure and the texture to serve as differentiating features. In this paper, we do a literature review of CNN based techniques for Image denoising, Image super resolution, contrast enhancement, improving images in low light conditions, and then propose a method utilizing the above mechanisms, and our own devised system which uses nearest neighbor classification to rebuild images, along with an HDR photography approach similar to Google RAISR's.