Detection and extraction of pantograph region from bank cheque images

Image matching and object extraction have been at the heart of computer vision research since the last two decades invigorated by the introduction of SIFT features by David Lowe. This paper proposes an approach for the same for its application to a specific problem, i.e. to extract the rectangular pantograph region from the cheque images used by banks in India. The proposed approach relies upon maximally stable extremal regions (MSER) for the keypoint localization and followed by speeded up robust features (SURF) for feature description. Then, these features from the template are matched with those in the cheque image. After the matching is performed, location of the detected blobs is identified and a binary image is generated with these locations as white and the rest of the background as black. From this binary image, the region of interest can be figured out by simple procedure based on iteration of a rectangular mask sized window throughout the binary image. The location having the largest density of the minority pixels is determined and this region is further extracted out from the initial input image. This is the first attempt of its kind to create an automated technique for extraction of the pantograph region from a cheque image. The results are encouraging. Also, this extraction of the pantograph region can be further applied on other images where it is being used to provide document security.