Special Topics in Image Engineering

Image restoration and enhancement problems have been treated as fundamental issues of image processing and computer vision. The image restoration estimates the original (clean) images from their corrupted and noisy image inputs in many forms of motion blur, noise and camera mis-focus etc. Therefore it is performed by reversing the degradation process that causes corrupted images. Different from image restoration, the image enhancement aims at improving the subjective perceptual quality of corrupted images, not necessarily producing the realistic data from a scientific point of view. In general, the image enhancement uses no a priori models of the processes that have created the images.
In recent years, the demand for ultra-high quality images has been increased with the advance in image restoration and enhancement. Especially, deep learning based approaches to image restoration and enhancement have made a great success in performance against the traditional approaches. In this class, we review the traditional approaches to image restoration and enhancement with the analysis of their limitation, and study very recent deep learning based methods using convolution neural networks, recurrent neural networks and generative networks. The students are exposed and experienced to very recent advanced methods in image restoration and enhancement via class lectures and homework assignments and terms projects.