A new image‐based unlevelness index

Abstract

A new quantitative unlevelness index based on the Fourier transformation frequency component is introduced for evaluation of the degree of unlevelness of a set of dyed fabrics with different surface colour uniformities. A series of dyed denims with different degrees of unlevelness were prepared, and the degree of uneven appearance of fabrics was ranked by a group of observers. The surfaces of fabrics were imaged by a conventional scanner, and the Fourier transform was employed to compute the spectrum of desired images. It was found that the low-frequency components of the computed matrix were stronger than the others, while its DC component, which related to the mean of the desired image, was too large. By this method, it was demonstrated that the fraction of the sum of the maximum of the second to sixth columns of the Fourier components of the captured image to the maximum of the first column component varied with the degree of unlevelness of the desired surfaces. The performance of the method was compared with five spectral and image based instrumental levelness–unlevelness indices, as well as those reported by visual ranking. Based on the results, the Fourier transformation method and the singular value decomposition technique show the best agreement with visual evaluation results, but the singular value decomposition method requires a longer computation time.