Effective Image Deblurring Based on Improved Image Edge Information and Blur Kernel Estimation
J Mercy, R. Lawanya, S. Nandhini, M Saravanan · 2022 8th International Conference on Advanced Computing and Communication Systems (ICACCS) · 2022
Image deblurring is one of the prevalent difficulties in image processing. Image blur is an unintentional loss of bandwidth that is difficult to avoid and affects image quality. Image deblurring can be done in a variety of ways, although most of them are restricted to motion blur and lack image edge information. An effective approach of image deblurring is presented to enhance the efficiency of image edge information and blur kernel estimation. In the luminance channel, the suggested technique employs high frequency layer information resulting from a input image using the 2D-haar wavelet technique. Following that, adaptive canny edge detection is performed, after that sliding window method is used in order to split the image into nine regions. The rich edge region index of each region is then used to extract the region with the most information about the edge. At the end, Convex kernel normalization and blind singular value decomposition are used to figure out the texture feature. The method is shown in MATLAB R2013a. Various quality factors has been measured to figure out how accurate it is. Efficacy tests reveal that the proposed strategy outperforms others.