Credit Card Processing Using Cell Phone Images

A new method to extract credit card information from cell phone images is described. The credit card image is preprocessed using Contrast Limited Adaptive Histogram Equalization (CLAHE) for contrast enhancement and non-linear soft coring technique for noise removal. The card bounding box is detected using Hough transform and geometric line processing, with iterative algorithm to improve robustness. The image is then morphed into a normalized template and the type of the card is detected using phase correlation. A novel method using feature detection and key-point projection is used to further segment the image into regions of interest (ROI). The ROIs are processed using an adaptive threshold computation technique which is important for obtaining reliable results from the Optical Character Recognition (OCR) system used to extract text information from the images. Multi-channel data processing from the color image is applied to improve the accuracy of the detection and an arbitration scheme is used to determine the extracted text. Experimental results demonstrate high hit-rate for number recognition, but show poor extraction of name field. Further work is needed before this system can be used in a commercial environment. Keywords-credit card processing, OCR, contrast enhancement, non-linear noise reduction