Algorithm of a Perspective Transform-Based PDF417 Barcode Recognition

Abstract

When a PDF417 barcode are recognized, there are major recognition processes such as segmentation, normalization, and decoding. Among them, the segmentation and normalization steps are very important because they have a strong influence on the rate of barcode recognition. There are also previous segmentation and normalization techniques of processing barcode image, but some issues as follows. First, the previous normalization techniques need an additional restoration process and apply an interpolation process. Second, the previous recognition algorithms recognize a barcode image well only when it is placed in the predefined rectangular area. Therefore, we propose a novel segmentation and normalization method in PDF417 with the aims of improving its recognition rate and precision. The segmentation process to detect the barcode area in an image uses the conventional morphology and Hough transform methods. The normalization process of the bar code region is based on the conventional perspective transformation and warping algorithms. In addition, we perform experiments using both experimental and actual data for evaluating our algorithms. Consequently, our experimental results can be summarized as follows. First, our method showed a stable performance over existing PDF417 barcode detection and recognition. Second, it overcame the limitation problem where the location of an input image should locate in a predefined rectangle area. Finally, it is expected that our result can be used as a restoration tool of printed images such as documents and pictures.