Segmentation for Fabric Pilling Images Based on Edge Flow

Pilling image segmentation is a key step in the objective evaluation system of fabric pilling level. The traditional methods of pilling segmentation apply to only one or more fabrics. In order to improve the properties of the evaluation system, an algorithm based on edge flow is proposed in this paper. The algorithm makes use of a prediction model to identify the image color or direction change of image texture in the given scale, and an edge flow vector is generated. The edge points of image are detected through the propagation of edge flow vector. The algorithm is proved to have strong adaptability and better effect for more fabric pilling images through the experiments.