| When machine vision is used to identify low contrast paper defects, if single edge detection algorithm is adopted, paper defects edge location become inaccurate, anti-noise performance is not good because of low contrast between the backgrounded and paper defects. This paper proposes a solution for this problem. Firstly, the paper containing low contrast paper defects should be edge detected by LOG operator and mathematical morphology operator separately. Then, the two images which are obtained by two types of edge detection algorithms should be merged by wavelet. In the fusion image, the paper defect’s edge location is accurate and that algorithm has the ability of anti-noise. Finally, the validity and dependability of the method and arithmetic have been validated. |
