Pilling Segmentation Technique In Fabric Pilling Objective Evaluation System

In the system of objective evaluation of fabric pilling, pilling image segmentation is a crucial initial step. The result of segmentation is associated with fabric pilling grade directly. While considerable researches and processes have been made by some people in home and foreign, but the generality of the algorithms of pill image segmentation on a large variety of textile image have not been established. One of the difficulties arises from the fact that most textile images are rich in the information of color and texture, and these features need to be integrated to improving the quality of the system of evaluation of fabric pilling grade.In this thesis, a scheme of pill image segmentation based on edge flow is discussed. This algorithm utilized a predictive model to identify the direction change in color or texture at each image location at a given scale, and construct an edge flow vector. By propagating the edge flow vector, the boundaries can be detected at image locations which encounter two opposite directions of edge flow vector in the stable stage. A scale factor that is also referring to the standard difference of Gaussian function is the only significant control parameter that is needed by the algorithm. The scheme facilitates integration of color and texture information into a single frame work for different scheme of edge detection. Also, experiment of results show that this method can be fit for most textiles in segmentation of pill image, moreover, the effect is satisfactory.