COLOR REPRESENTATION OF TEXTILE FABRIC USING MULTISPECTRAL IMAGING TECHNOLOGY

As the normal colour measurement method of using a spectrophotometer is impossible to measure colours of multi-coloured textiles fabric samples such as color printing sample, and knitted samples using colored yams, color measurement method based on the multispectral imaging technique is a viable alternative. In this paper, we propose a new novel approach to represent the spectral colors of these textile samples. Firstly, the reflectance of captured image by a self-developed multispectral imaging system is recovered by the Wiener estimation method at each pixel spatial location. Secondly, a bilateral filter method is used to remove the signal dependent noise mainly caused by the imaged object during the imaging acquisition procedure. After that, an unsupervised tree structure is constructed to segment denoised image. Finally, all color regions are extracted from the segmented image of a multi-color printing sample to represent the dominant colors weighted with the segmented block size. In the case of representation of yam-dyed textile image, a K-mean clustering method is further used to extract the spectral colors of weft and warp yam, respectively.