Sparse signal processing for image applications

Abstract

Image processing is a popular and well-researched topic in the signal processing area, and image denoising and inpainting form the cornerstone of image processing. Since there are various ways to denoise noisy images or inpaint images with missing pixels such as deep-learning-based methods, the approach applying sparse signal processing techniques is still worth the attention because it exploits the intrinsic characteristic of sparsity in images. In this dissertation, the K-SVD algorithm combined with the Orthogonal Matching Pursuit (OMP) algorithm is explored and applied in image denoising and inpainting. Experimental results show that this approach can effectively improve the visual quality of images and reduce flaws in images.