Reliability of statistic evaluation of microscopic pictures taken from knitted fabrics

Abstract

One of the techniques which can be used to quantitatively evaluate images statistically is the so-called random-walk approach. The resulting Hurst exponent is a measure of the complexity of the picture. Especially long, fine elements in the image, such as fibres, influence the Hurst exponent significantly. Thus, determination of the Hurst exponent has been suggested as new method to measure the hairiness of yarns or knitted fabrics, since existing hairiness measurement instruments are based on different measurement principles which are not comparable. While the principal usability of this method for hairiness detection has been shown in former projects, the absolute value of the calculated Hurst exponents depends on the technique to take the photographic image of a sample, to transfer it into a monochrome picture, and on possible image processing steps. This article gives an overview of edge detection filters, possible definitions of the threshold value between black and white for the transformation into a monochrome image, etc. It shows how these parameters should be chosen in case of typical textile samples and correlates the challenges of this novel method with well-known problems of common techniques to measure yarn and fabric hairiness.