Restoration of Blurred Images and Video Sequences

Abstract

In our day to day life high quality digital images have become ubiquitous. How- ever, these images are always degraded by some uncertainty and imprecision between object and measurement system. In the study of image processing, image restoration is very important aspect as it is method to compensate degradation of the images. This thesis deals with the blurred images by relative motion or defocus objects spe- cially when the imaging system is in moving vehicles or held by human hands. The thesis work handles the restoration of blurred images and video sequences. This thesis has been bifurcated in two parts, where the first part comprises of method to restore motion and defocus degraded images while second comprises of video deblurring. The first part deals with identification of Point Spread Function (PSF) and estimation of original image by using this estimated PSF and the degraded image. In this thesis, PSF estimation is done using patch-based method and kurtosis & skewness based techniques. In second part the main concentration is on video deblurring, which is achieved by two techniques namely using hough transformation and kurtosis & skewness based techniques. Video restoration process is done by estimating the blurring parameters from a video first and then try to restore it by performing different and appropriate methods. Compared to hough transformation based technique, increment in PSNR from 16.25 dB to 31.94 dB is achieved using kurtosis & skewness based technique.