Please see final paper for
full report.
Or email 'sauravc @ cs.washington.edu' for futher inquiries.
Abstract:
Currently there is a lack of programs for decoding bar codes in images taken by camera phones (cell phones equipped with cameras). The majority of camera phones generate low resolution (680x420) pictures that are noisy, unevenly lit, radially distorted and askew. Although packages exist that can read bar codes from higher resolution images, most if not all of these are proprietary and are not designed to deal with such low resolution and high noise. We have come up with several novel approaches that utilize and extend Hough transforms and homographies to automatically detect and remove distortions within an image. We also have also come up with a robust algorithm to decode the barcode once it is located within the image.(June 2004)