Research On The Aesthetic Quality Evaluation Of Chinese Handwritings

With the rapid development of image acquisition and visual computing technology,various information and data pour in.Since the image is intuitive and vivid,it is becoming more widespread use of information expression form.Computer-aided aesthetic evaluation is one of the most popular topics in the field of artificial intelligence.In Chinese language education in primary and secondary schools,Chinese characters have begun to enter the syllabus.Based on the extensive review of the existing literature,this paper focuses on the aesthetic evaluation of Chinese handwritings,the main works of this paper are as followes:(1)The preprocessing of Chinese handwriting image is studied.The method of using edge preserving filtering is proposed for denoising,and experiments show that this method can remove the impulse noise effectively,and will not cause edge fog of the handwriting image.In the process of binaryzation,OTSU,an adaptive method,is proposed,which can not only be simple and effective,but also can avoid the accident of selecting different handwriting samples.(2)The aesthetic evaluation of Chinese handwritings based on characteristics of radical is studied.Global features such as the convex hull,the minimum bounding box are focused on to be quantitatively analysed.BP algorithm is used for training so that the machine scoring of each Chinese handwriting sample can be calculated according to the certain rules.By comparing it with manual scoring results,the method can be proved to be feasible.(3)The method of the aesthetic evaluation of Chinese handwritings based on PCA and SVM is proposed.Firstly,PCA is used to extract the main feature of the Chinese handwriting images,and then,the whitening matrix and reconstructed images can be obtained.Secondly,SVM is used to multiply classify the Chinese handwriting images into excellent,good,medium and poor such four evaluation grades.The advantages and disadvantages of various SVM multi-class classifiers are analyzed and compared,and the "one against one" construction method is selected.Then,different models and kernel functions of SVM are compared by grouped experiments.Thus,the C-SVM model with RBF kernel function is proved to be valid in this paper.(4)According to the obvious characteristics of the stroke texture of Chinese handwritings and the advantage of the Gabor transform for the local texture feature extraction,the method of the aesthetic evaluation of Chinese handwritings based on Gabor transform and S VM is proposed.Firstly,the selection and significance of Gabor filters with different parameters are studied.Combined with the characteristics of Chinese handwritings of using horizontal,vertical,left-falling and right-falling strokes as main strokes,appropriate direction parameters are selected to extract the texture feature of Chinese handwritings by Gabor method.Secondly,PCA is used to extract the main features of Chinese handwritings after Gabor filtering to achieve the goal of dimensionality reduction.Finally,different groups of training samples are selected,and the C-SVM model with RBF kernel function is used to make the level evaluation classification of the samples,whose effect on the experiment performance in this paper is verified by comparison.The experimental results show that the Gabor method is more effective for the aesthetic evaluation of this paper.In this paper,traditional methods are applied to the experiments and discussing of the new problems of Chinese handwritings,and it also becomes the characteristics and innovations of this field.