A fast algorithm for detection and removal of punch holes from scanned document images

A fast algorithm is proposed in this paper for detecting and removing a special circular noise caused by punched holes from scanned document images. By elaborately reducing the original image, the algorithm can fast detect and locate all of punched holes in the image. To avoid false detections, the algorithm validates the holes noises in the roughly located regions through Hough transformation. Punched holes are finally removed by fitting a bi-linear blending Coons surface which interpolates along the four edges of the noisy region. Our experiments show that the algorithm described here is efficient and robust, and the average runtime is much less than the normal scanning time thus well suitable for real-time processing.