Finding and reading barcodes using low resolution camera has increasing demands in mobile phone, automatic freight handlings, POS(Point-of-Sale) systems, parts assembling, etc. As an image analysis, barcode detection may be simpler than edge detection...
Finding and reading barcodes using low resolution camera has increasing demands in mobile phone, automatic freight handlings, POS(Point-of-Sale) systems, parts assembling, etc. As an image analysis, barcode detection may be simpler than edge detection because bars are parallel, and their bars and spaces between them are multiple of unit bar width. However, the accuracy and efficiency of edge features deteriorates because of the small and detailed barcode patterns which have bars and spaces in an image. So, finding barcodes using a region-based analysis of low resolution image is more desirable. As the method of barcode extraction, the proposed algorithm computed covariances by using Consistent Gradient Operator proposed by Shigeru ANDO, and then categorized image into UNIVAR(Unidirectional Variation), OMNIVAR (Omnidirectional Variation) and NOVAR(No Variation) based on variation of directionality.
Binarization is very important problem when we recognize barcode. If the binarization result were poor, the recognition would be difficult or even unavailable. From this point of view, the proposed algorithm used adaptive binarization based on local thresholding which updates the threshold value whenever the Laplacian sign of the input image changes. According to the simulation, the proposed algorithm achieved performance improvement about 5% compared to the conventional method.