Image Inpainting Based On Sparse Representation with Dictionary Learning and Construction

The practice of imperceptibly altering images has a long history in artistic production. In the field of image processing, there have been significant advancements in studying image deblurring, denoising, and inpainting, contributing to picture enhancement and restoration. These studies have garnered widespread recognition due to substantial improvements in recent years. One intriguing area of research in picture restoration is the reconstruction of missing regions, which has various real-world applications. This paper focuses on removing extraneous text material from images through an image inpainting task. The K-SVD approach is utilized to acquire sparse representation dictionaries from the given image. Random and DCT dictionaries are constructed for image processing, and through the application of the K-SVD technique, the dictionaries are trained to adapt effectively to image patches with different properties.