Study on Detection Method of Low Contrast Fabric Defects Image Based on Regional Variation Saliency

In order to solve the problem of low contrast fabric defect detection accuracy, a new detection method based on regional variation saliency was proposed. Firstly, the cartoon texture model is used to filter the fabric image to remove the yarn texture. Then, the texture features of the image block are extracted to construct a visual saliency map to highlight the defect area. Among them, the texture features are calculated based on the regional variation of the image block. Finally, the threshold segmentation algorithm is used to segment the visual saliency map and obtain the defect area. The results show that this method can detect the defect information in the image of low contrast plain or twill fabric more completely than the existing methods.