Shadow removal from document image using normalized convolution smoothing and plane approximation

In this paper, we propose a shadow removal method for document images that preserves natural appearance. Although several shadow removal methods have been proposed, these methods have the following problems: some of the shadows remain without being removed, character degradation, and time-consuming. In proposed method, character regions are first extracted and then removed by normalized convolution smoothing to estimate the background image. Next, shadows in the image are removed using this image, the background color obtained by plane approximation, and the input image. The effectiveness of the proposed method is validated through comparative experiments.