Integrating feature descriptors to scanpath-based decoding of deformed barcode images

Abstract

There is an extensive literature on localization and decoding of barcode images. Most traditional scanline-based methods have difficulty decoding a highly deformed barcode. This paper thus presents a scanpath-based method to manage the decoding of deformed barcode images. The notion of local orientedness based on HOG descriptors, called Modal Orientation Deviation (MOD), is introduced to help in scanpath construction. The effectiveness of the proposed method is also discussed according to the results of tests reported.