Periodicity Estimation of Regular Textile Fabrics Based on Energy Function

Abstract

The periodicity of the fabric pattern is one of the most important characteristics of textiles. It has a very significant impact on the structural analysis and practical application of textiles. In this paper, a method to determine the periodicity of both patterns and textures is proposed. For a given periodic texture image, we first transform it to HSV form from RGB. Then we derive the component of H and put a histogram equalization process on it. After that, we use the energy cues computed from the sum and difference histograms of H image. When the variable parameter in the energy function matches with texture element size or its integer multiples in both horizontal and vertical direction, the energy function reach its local maximum, then we can automatically obtain the texture element size according to this phenomenon. We synthesized a new image with the obtained texture element and compare it with the original image to evaluate the performance of our method, the experimental results show good performance of our method.