Strong Shadow Removal of Text Document Images Based on Background Estimation and Shading Scale
Bingshu Wang, Shuang Feng, C. L. Philip Chen · 2020 7th International Conference on Information, Cybernetics, and Computational Social Systems (ICCSS) · 2020
Shadows may bring uncomfortable perception when taking text document images. Previous work mainly focuses on weak shadow removal issue. However, strong shadow removal is still a challenging task. This paper proposes a method based on background estimation and shading scale to remove strong shadows from text document images. Firstly, background color estimation is designed by a number of iterations through neighboring pixels' propagation. Then, umbra and penumbra are separated by morphological operations and processed by a divide-and-conquer strategy. For umbra, shading scale strategy is exploited to obtain unshadowed result. For penumbra, background replace strategy is designed to remove the shadow regions. Finally, a reference background and a text binary image are combined to generate unshadowed image. Experiments conducted on some typical strong shadow images demonstrate the effectiveness of our method.