Yarn separation for structure analysis of textile fabric on three-dimensional CT image based on estimated filament direction

In this paper, a yarn separation method is proposed for structure analysis of textile fabric on its three-dimensional (3D) image obtained by an X-ray computed tomography (CT). In this method, the directions of the filaments, which a yarn consists of, are firstly estimated at all the voxels of the 3D image by correlating the voxel values with a filament model. Each filament is secondly reconstructed by connecting the voxels of the filaments based on the estimated filament directions. Each yarn is then reconstructed by binding up the reconstructed filaments using a newly defined distance between two filaments. Consequently, each yarn of the textile fabric on the 3D image is separated. The effectiveness of the proposed yarn separation method is discussed by experimentally applying this method to 3D image of a plain knitted fabric.