Digital Image Inpainting by Estimating Wavelet Coefficient Decays From Regularity Property and Besov Spaces
Sérgio R. M. Penedo, Flávio A. M. Cipparrone, João F. Justo · IEEE Access · 2018
Inpainting methods play a major role in image processing, minimizing the effects of data block loss in image transmission. They analyze primarily the spatial correlation between the portions of an image, usually without explicitly exploring information on the frequency domain. Techniques that evaluate losses of blocks in images do not locally describe a figure of merit for consistent measurement of spatial correlation between pixels of each corrupted block. This investigation presents a digital image inpainting technique, extensible to video applications, using the property of wavelet regularity in multiresolution image analysis, described locally in the sense of Besov vector spaces. Their generalized smoothness allows widespread changes in wavelet coefficients of image data, concealing errors in visual information. This technique estimates decay curves of wavelet coefficients from pixels, based in the observation of the regularity property, correcting pixel blocks contaminated by noise. This investigation proposes an algorithm in the wavelet transform domain, which detects and fixes damaged pixel blocks from a designed decay function for wavelet coefficients. Simulation results, obtained in corrupted images by several noise patterns, indicated better performance of this methodology when compared with other ones.