The Simulation Of Board Carving In The Digital Synthesis Of Out-of-Print Woodcuts

Yunnan out-of-print woodcuts is the typical representative of Yunnan local minorities’characteristic painting. It combines the art of painting, carving and printing. The artist use one board to print several layers of color on the paper. This usually involves carving a small part of the board away, and printing the board to the paper. This process can be repeated many times over. Once the artist finishes the work, the board is destroyed and no more prints can be made. This is why it is named out-of-print. The digital synthesis of Yunnan out-of-print woodcuts is a challenging work, which consists of synthesizing score texture, generating digital board and simulating the color printing.This thesis focuses on generating digital board of Yunnan out-of-print woodcuts, which is one of the three parts in the synthesis of Yunnan out-of-print woodcuts. Firstly, we give a brief description of the actual creation of the Yunnan out-of-print woodcuts, which is the physical foundation of the digital synthesis work. Secondly, by consulting the process of the out-of-print woodcuts, we introduce a digital board generating framework, which involves image abstraction, features extraction and image segmentation. This framework contains two stages.In stage one, we use a PDE-based nonlinear diffusion filtering to perform an image abstraction and extract the color features. After that, the texture features are extracted with nonlinear structure tensor. Combining the color features and texture features, we obtain a5-D features vector. This5-D features vector will be used. in the stage two.In stage two, we introduce an image multi-class segmentation method to generating the digital board of out-of-print woodcuts. Three statistic models are introduced in this thesis:(1) Standard Gaussian Mixture Model;(2)MRF model defined on the class label;(3)MRF model defined on the parameters of priori distributions. After that, we use the third model to perform the image segmentation. The5-D features vector obtained in stage one is used as the input data set of the segmentation algorithm that results in labelled image pixels. By substituting both color features and texture features to the model, we get competitive segmentation results.Finally, we generate a series of binary images on the basis of the image segments. Each binary image describe the shape of the board, thus the series of binary images can be used to simulate the shape sequence of the board in the process of out-of-print.