Abstract: For the correct localization of the two-dimensional bar code in the image, the paper proposes a multi-scale segmentation location algorithm. In the first segmentation, the image is divided into several blocks, then, according to the black and white pixel ratio, variance of high-frequency sub-band coefficients, the blocks would be classified into bar code and non-bar code areas. By gradually increasing the scale, the block is divided into four sub-blocks, the segmentation result is recursively refined at each scale based on the multi-scale segmentation model. And so on, until the finest scale, the location of bar code area in the image is gotten. Experimental results show that the algorithm can be better to locate the bar code area. Especially in the case of complex background with large amounts of text, this method has certain advantages.

Key words: two-dimensional bar code location, multi-scale segmentation, feature extraction, black and white pixel ratio