Yarn Count Estimation in Woven Fabrics Using FFT-Based Image Analysis

Yarn count is a primary parameter in fabric structure that greatly influences fabric quality, usage, and design. It is defined as the ratio between the length and weight of the yarn. The Ne unit belongs to the indirect count system (constant weight, variable length) and is commonly used for cotton or cotton-blended fabrics. Conventionally, yarn count is measured manually by determining the yarn diameter, but this method often results in inaccuracies. An automated analysis using Fast Fourier Transform (FFT) can extract directional and repetitive patterns from grayscale texture images, and is therefore applied for automatic yarn count estimation. In this study, the accuracy of yarn count measurement is improved by applying linear regression, resulting average error for plain weave fabrics (5.51% for warp and 14.22% for weft), followed by twill weave (9.72% warp, 24.00% weft), and satin weave (13.94% weft). However, a high error was observed in the warp yarn of satin weave (31.42%), which is attributed to visual distortion in the yarn appearance.