An Automated Color and Texture Image Analyzing System for Embroidery Fabrics

• Main Authors: Jung-Hsiang Cheng, 鄭榮祥
• Other Authors: Chung-Feng Kuo
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/29115936761906496335

碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 103 === This study created an automated color and texture image analyzing system for embroidery fabrics. First, a scanner was used to obtain the digitized color image of the embroidery fabric in RGB mode. Then the image was converted to a digitized color image in CIELAB mode. In this stage, noise reduction methods such as mean filtering, median filtering, one-time and two-time wavelet transformation processing on lightness layer of CIELAB color space were conducted. The Peak Signal to Noise Ratio (PSNR) method compared the digitized image quality status between the original to the de-noising image. In the results, one-time wavelet transformation de-noising processing improved the digitized image smoothing with low processing loss. The pre-processed embroidery fabric images were developed with the one-time wavelet transformation. The Fuzzy C-means (FCM) clustering method was employed to run color separation and regional separation. The main texture analysis of embroidery fabrics is based in Entropy and Hough transform approach to main texture feature is obtained. The yarn texture analysis was based on Gabor filters approach to obtain yarn texture features. Pre-processed embroidery fabric images used a set of Gabor filters with different frequencies and orientations for extracted yarn texture information. The orientation parameter for Gabor filters was one algorithm of texture recognition on directional fields. Using three scales and three directional convolution operations, the Gabor filters detected gray level yarn texture information of the embroidery fabric. The appearance of the yarn texture feature was found with automated threshold processing. It calculated the optimal threshold through the Otsu algorithm and executed binary processing. Finally, the result was   achieved by thinning method processing, which removed the outlier to obtain yarn texture feature. The results of this study confirmed that the method could be applied to embroidery color and texture analysis.