Retrospective tracking for barcode reading

A novel method for fast reading barcodes commonly encountered in industrial applications is presented. The method mainly consists of two stages: (1) a process characterizing in cross scanning an overly downscaled image to extract candidate barcode skeletons is used to locate barcodes in an image, wherein at least one skeleton is extracted from the downscaled binary image; coordinates of at least one point of each skeleton are enlarged back to their original resolution and used as center points of plural detection windows in the original image plane. (2) a Sobel-based grading scheme is employed to determine wining detection windows for reading barcodes, wherein a retrospective tracking mechanism is proposed to deal with warped, slanted or even stained barcodes. Since the center points of the wining detection windows are directly treated as sampling points of the barcodes, pixel by pixel search is dispensable and computation time can be greatly reduced.