A Statistical Approach for Multi-frame Shadow Movement Detection and Shadow Removal for Document Capture
Prasenjit Mondal, Ankit Bal · 2022 30th European Signal Processing Conference (EUSIPCO) · 2022
Shadow detection is an integral part of scanned document image processing. In this paper, we have introduced a new technique for identifying shadow movement regions in documents. Provided a sequence of image frames, our algorithm identifies documents from an image using a boundary detection algorithm followed by computing their respective perspective corrected images. Further, intensity map images are computed from the perspective corrected images. Finally, a statistical analysis is carried out on the different blocks of the intensity maps to estimate the shadow movement regions. The identified blocks along with a shadow removal algorithm are used to remove shadows from the documents. The proposed algorithm has been tested in with the Adobe Scan application and the experimental results show the efficacy of the proposed algorithm. The algorithm has a low computation complexity which makes it suitable to use in different mobile applications.