Pills on the surface of a fabric are an undesired appearance defect caused by wear, particularly for knitted fabrics. Pilling degree is evaluated by comparing the standard pilling images according to EN ISO 12945-2:2000 in laboratories with the samples obtained from processed fabrics in different devices after artificially creating pills. Evaluation by standard images subjectively leads some confusion and problems between different laboratories. Objective evaluation approaches using image processing algorithms are actively researched instead of subjective evaluation. In order to develop an objective evaluation system compatible with the standards used in industry using image processing methods is the main aim of this research. Segmentation partitions an image into distinct regions, namely distinct objects which refer to the pills and background in this work. We present the results of thresholding methods used for segmentation of pills on the surface of standard images which is the first and one of the most important step in order to develop an objective system. If the pills cannot be segmented correctly, the rest of the system cannot work satisfactorily. Although intensity thresholding is the oldest approach for segmentation, it is still used and researched extensively. The performance of thresholding methods is researched and the local thresholding approach is studied for pill segmentation.
Anahtar Kelimeler: Pilling Grade, Pill Segmentation, Local Thresholding, Image Processing
|
