Extracting Reflection Components of Glossy Woven Fabrics with High Dynamic Range Image

This paper describes a method of estimating the color vectors of the body and the interface reflection component of glossy woven fabrics based on the color histogram analysis of a local region. We choose the homogenous colored polyester satin as objective glossy fabrics. We fix woven fabrics on the surface of a side of cylinder and acquire images, which consist of both regions with highlights and with no highlights, with different amount of exposure. We then generate a high dynamic range image from a set of acquired images of multiple exposure time. In the local region, the body reflection is constant but the interface reflection varies with the microscopic roughness. Therefore, the distribution of the color cluster of the local region is linear. We estimate the color vectors based on the linearity of the color cluster distribution of a local region. We separate the body and the interface reflection component from the original image of the objective polyester satin by means of the estimated color vectors.