The Study For Chinese Calligraphy Identification Based On The Style Characteristic

It has a profound significance to protect and carry forward the traditional Chinese art and culture, and to help the development and promotion of computer art, which based on the image feature extraction of Chinese calligraphy identification research. At present, it mainly relies on the connoisseur to discern the false from the genuine by subjective experience in the field of art of calligraphy identification. However, it is difficult to complete by subjective experience for a large amount of data. How to provide an effective identification criteria for calligraphy through computer? This article starts from five aspects which are acquisition of the original works, word pretreatment, extraction of calligraphy stroke, extraction of individual character style feature, identification of calligraphy, and finally gives the probability of the calligraphy for genuine.For more traditional skeleton extraction algorithm of burr and intersections are prone to the problem of distortion, this paper puts forward a new method of handwriting character skeleton extraction. In thought, on the basis of shape decomposition, skeleton extraction of calligraphy character will be turned into a step-by-step problem. Because of the point cloud model can make full use of the existing Chinese characters of the underlying information, so this article use of point cloud model, the k-means clustering algorithm are adopted calligraphy word segmentation images;And put forward a kind of suitable for characteristics of calligraphy words in neighbourhood domain pixel distance incremental clustering algorithm for clustering center skeleton extraction calligraphy word parts. Finally it based on the character of skeleton to form a complete skeleton topology connection parts.Calligrapher, through years of practice, will have their own style when writing, and these features are stable, which caused by long-term different writing habits. In the view of this characteristic, this paper uses gaussian distribution model to extract multiple feature points of the calligrapher’s writing, according to the different rules from the feature of calligrapher level and structure level. Then select and estimate the calligrapher’s feature point and weight, constructs the individual character style characteristic vector, and set up the model of calligrapher’s characteristics for each feature point by using the model of a gaussian distribution.In the feature matching stage, this paper proposes a composite algorithm based on geometric features and frequency domain correlation matching algorithm for feature matching in order to do next step calculate on similarity, which based on the classical Fourier-Mellin transform method of feature matching algorithm. Using each diagnostic function of style characteristic vector to identify, and to conclude the probability of each feature point for genuine, then to calculate the probability to determine every work for genuine. At last, use the Visual Studio 2010 as a platform for system implementation.