Image deblurring using machine learning models

Abstract

In photography, images can appear blurred for a number of reasons including lens defects, camera motion, and camera focus. This work investigate the deblurring of images using deep learning techniques. The chosen approach is to use a deep stacked hierarchy using a multi-patch network. This will allow one to identify how the deep learning of processing blurred images can be improved to generate better deblurred images. We also investigated whether the deblurring process can be done faster and without losing image quality. The projects assesses the theoretical and practical implications that can be used in digital image processing, while focusing blind image deconvolution aspect of processing. This “blind” aspect in image processing refers to a particular function known as the PSF that was involved in blurring the image. The PSF is assumed to be unknown and hence why it is called ‘blind’. Image deblurring is done on images selected from publicly available datasets. These newly deblurred images will then be compared to those deblurred other existing image deblurring functions. As the area of this project is data science, the programming language to be used is Python as it is designed to aid in data analysis and visualization with special features, including external libraries. This will result in standard images to have enhanced quality once the program finishes execution and allows one to compare the original image (the ground truth) to the new and improved image. The evaluation of the system is based on test images which have not been used for training to determine if image can be improved by the trained network. The details of both images will be matched side by side to show that the image has indeed been deblurred and enhanced.