Abstract: Four self-adaptive threshod algorithms were employed, respectively, to detect defects on plain and twill fabrics based on images and the results of the detection are analyzed, so as to compare the detection effcet of the four adaptive threshold algorithms in detecting defects of the fabercs whose gray values have a big difference from the background and then to find out their advantages and disadvantages. The experiment resultshows that all the four algorithms can be successfully used in fabric defectdetection. The order of the detecting effcet is: local threshold segmentation> improved Otsu> Otsu> maximum entropy when the time consuming is not considered. However, when the time consuming is considered the order is: local threshold segmentation> Otsu> maximum entropy> improved Otrsu.

Key words: self-adaptive threshold segmentation, Otsu, improved Otsu, maximum entropy, local threshold segmentation