Image analysis of standard photographs of pilled fabric

An objective method for pilling evaluation in fabrics is developed from the image analysis of a set of standard photographs in this paper. The procedures involved in using the Fourier transform to remove fabric texture which interfere with pill recognition are demonstrated. The filtered image is segmented by a threshold under the assumption that the histogram can be simulated by Gaussian distribution. Pilling features including number of pills, area of pills, contrast and pill density are described and extracted. Three features are selected and normalized to be criterion for objective assessment. The grade of fabric pilling is assessed by use of the nearest neighbor algorithm. The experimental results of wool and blend fabric sample show the objective method is effective and feasible.