Base color recognition by tetragonal regression for overlapped watercolors

In this paper we propose a base color recognition method for multicolor printing with watercolors including some areas overlapped each printed layer. Watercolors have an important feature that the observed colors are different by the printing order. The proposed method is based on a particle density model to express the features of watercolors. The approach of this method is to solve an inverse problem, that is, determining a printed area corresponding to a watercolor from an observed image. Our purpose in this study is to estimate woodblocks from a woodblock print. We focus on multicolor woodblock prints such as Ukiyo-e: Japanese traditional woodblock printing, artistic documents. The base colors forming the given image are derived by planar regression bounded by a tetragon in the CIE L*u*v* color space. The tetragon is determined by the color distribution of the given image. The results of the base color estimation can be utilized to preserve techniques and prints of Ukiyo-e. An Ukiyo-e print is created from several woodblocks with several watercolors, and has a mixing color expression similar to the effects of watercolor painting.