Digital document image restoration using a blind source separation method based on copulas

Abstract

Abstract In the last few decades, digital image degradation issues, such as blur and noise due to the scanning process or the presence of spots, underwriting, overwriting or bleed-through/show-through effects on the image’s background has been a popular research field. To solve this problem, many background removal methods has been introduced in the literature which are based on local or adaptive filters in order to deal with the low-contrast issue. For this paper, we will be focusing on the bleed-through/show-through effects, which is already resolved in literature by an analogy between the front-ground and the background of the image, that is to say, a recognition of two images is required. To fix that problem, we suggest a new restoration method using blind source separation based on copulas theory that models the dependency structure, with the aim of improving text readability and OCR efficiency.