Blind Image Deblurring using GLCM and ElasticNet Regularization

The well-known source of digital degradation is camera shake, photos under dim light and a handheld camera etc. Extensive research has taken place over the last decade in the field of retrieving a latent image from burry input; most of them work quite well, but very often incur to blur in edges. This paper has been proposed a new deblurring method in which the high-frequency layer is extracted from the blurred image using a 2D Haar wavelet transform in the luminance channel, then from the high-frequency layer, rich edge region is extracted using GLCM and sliding window concepts after the canny edge detection process. Finally, the extracted rich edge region is used to estimate the blur kernel using the elastic net regularization of singular value. Here regularization is used to avoid over-fitting of the data and reduces the blurring effects of the image. Experimental result demonstrates that the proposed deblurring algorithm achieves the better results on natural images which are evaluated using the parameter such as PSNR and SSIM.