| 摘要 |
<p>A method wherein an image (I) of an object bearing an optical code is divided (120) into a number of elementary images (If) for which a brightness gradient vector (G) is calculated (130). The calculated gradients (G) of a magnitude above (140) a given threshold value (Glim) are transformed (190) to eliminate the information relative to the direction of the vector, while maintaining the information relative to the path of the vector. The transformed (190) and quantized (200) vectors of a subimage (Ip) including a number of elementary images (If) are added (220) to calculate a total gradient vector, which is compared (230) with reference values to select (240) significant subimages containing a sufficient number of gradient vectors of substantially the same path. <IMAGE></p> |
