Chinese Journal of Computers   Full Text
Title	Statistic Model-Based Simulation on Calligraphy Creation
Authors	DONG Jun1) XU Miao1) PAN Yun-He2)
Address	1)(Software Engineering Institute, East China Normal University, Shanghai 200062) 2)(Artificial Intelligence Institute, Zhejiang University, Hangzhou 310027)
Year	2008
Issue	No.7(1276—1282)
Abstract &   Background	Abstract Calligraphy is classic course of imagery thinking. The authors manage to create new stroke shape of calligraphy character based on a statistical model with parameter: X=M(b). Where X is target stroke with new styles and parameter b is a vector to adjust the stroke features (e.g. long or short of the stroke). When the parameter is changed with some specific values, new style could be generated. Therefore, some samples are selected in order to form a training set. Each of them is marked with several points, including the points on the contour and the control points which will be used by generating the consecutive contour curve later. After that, the training set is aligned by using Generalized Procrustes Analysis and the primary features of these samples are extracted by Principal Components Analysis. Later on, a few characteristic features in the training set are extracted and represented by the form of several eigenvectors. With the extracted features, new styles of strokes can be generated. Then, stroke one(Horizontals) in "Li" style of Chinese calligraphy is illustrated with simulation results. Simulation of intelligence, patterns of imagery thinking and model of cognition are key issues here. Current work is only the first step. Keywords calligraphy creation; imagery thinking; statistics model; stroke; simulation Background Imagery thinking and abstraction thinking both are very important concerning thinking science, cognitive science as well as artificial intelligence. Whereas, there exists extraordinary unequal understanding between them while too less result on imagery thinking is ready because of no enough experiment evidences from neuroscience and much difficulty for field experts such as artists to utter their sensation and experience during their creation. To solve the problem, simulating imagery thinking on the basis of expert experience is necessary. Calligraphy creation is a classic course. In most cases, the automatic generation of Chinese calligraphy can be divided into several phases: (1) shape decomposition, (2) model creation and (3) artwork generation. An algorithmic framework for an advanced virtual brush to be used in interactive digital painting has been proposed. Compared with other virtual brushes, this system was designed to present a realistic brush in the sense that the system accurately and stably simulates the complex painting. Moreover, an intelligent system using a constraint-based analogous-reasoning process was presented by Xu Song-Hua et al. in 2005. The system fuses knowledge from multiple sources to support a restricted form of reasoning. The result is pleasant since it can automatically generate Chinese calligraphy that meets visually aesthetic requirements. However, the transformation on each level depends on the weight of different samples. The efficiency is a bit low since a lot of experience is required to adjust the weight. It would be better if some primary features of a character can be extracted at first. This is the main purpose of our current work, i.e., obtaining an approach to extract some critical features from large numbers of samples and then build a model with them. Now we can modify the stroke shape easily as the real creation course. The work has several potential applications. First of all, it can be applied in the publishing industry, especially those publications of ancient artworks. For some calligraphy works suffer a lot of abrasion during long history, it is impossible to maintain the original artwork. This approach could be used to generate some similar characters to be selected for publishers. Another application is to generate personalized fonts according to users’ preference. For example, users provide the system with their handwritings, based on which, some plausible styles could be generated. Users can adjust and choose one or more styles as their favorite appearance. The research is supported by Shanghai Basic Research Key Project (06JC14058), NSFC(30570584), National Basic Research Program of China (2005CB321904) etc.