Investigation the effect of weave repeat on woven fabric texture characterization using sparse representation

Abstract

A novel approach based on sparse representation is presented to investigate the impact of weave repeat on the characterization of woven fabric texture. Firstly, the test samples were represented by over-complete dictionary in the least squares sense. Secondly, the two indexes--PSNR value and RMSE were introduced to evaluate the representation performance. Thirdly, the image entropy was used to quantify the fabric surface texture, and then the samples were categorized according to the two indexes. Experimental results showed that our algorithm can approximate fabric texture very well and weave pattern has great impact on fabric texture reconstruction. Eight kinds of weaving patterns are classified into three categories, of which the basket fabric shows the best performance. The findings may be helpful in classification and automatic inspection of woven fabric texture.