Research Into Chinese Envelope Image Segmentation And Destination Address Block Location

The automated processing and sorting of letters and packages is an important task for modern mail delivery systems. Nowadays, postal automation has been integrated into the research field of digital image processing, computer vision and pattern recognition, since images of envelopes and parcels can be acquired and stored easier and cheaper than a decade ago.Segmentation of envelope images is the first stage of the postal automation, which is still a challenging problem due to the large variety of stamps, backgrounds, written text of the address (e.g. handwritten, printed, locations). Furthermore, the system must be able to precisely locate the address block in real time before send it to the OCR program. Since character recognition systems are trained to recognize only words and characters, any extraneous information can easily confuse them. Thus, successful address block location is a prerequisite for postal automation.This thesis firstly introduces the research status of envelope image segmentation and destination address block location. Then the background and several thresholding algorithms are illustrated in detail. Meanwhile, research and application in cell image segmentation by using a multi-stage thresholding approach is introduced. Next, the improved move-window robust automatic thresholding selection (RATS) algorithm is applied to threshold the camera-captured Chinese envelope images. Also, skew detection and correction algorithm is used to correct the result binary images, and opening operators in mathematical morphology are employed to extract and remove the rectangular frames containing postcode. Finally, a novel approach for locating the postal address block on Chinese envelopes base on support vector machines (SVM) is introduced.Over 800 envelope images are tested, including handwritten and machine-printed envelopes. Experimental results show that the proposed method can not only be effectively applied to machine printed Chinese envelopes, but also works with handwritten Chinese envelopes.