Classification Of Chinese And Western Painting Images Based On Brushstrokes Feature

Chinese and Western painting is one of the important forms of cultural works in the history of world civilization.In the process of history,there are many excellent paintings,which are important carriers of the study of history,culture,art and technology.With the increasing popularity of digital Internet technology,more and more paintings are displayed in the electronic library.It is easier for art lovers to appreciate excellent paintings.The increasing number of digital painting images brings scholars rich research resources.At the same time,how to effectively classify large-scale digital painting images is one of the heated issues to be solved.Differing from ordinary natural images,paintings have unique styles,techniques and artistic emotions.Brush stroke is one of the most basic painting elements reflecting these characteristics,and it also has very important visual features in painting research.This paper classifies Chinese and Western painting images and Western style painting images by extracting the unique stroke features of painting works.1)This paper presents a new method of stroke feature extraction.Firstly,Sobel operator is used to extract the stroke features,and a 3 * 3 filter is used to operate on the painting image aimimg to obtain the edge lines of the image.Because the strokes in the painting works are consistent,the morphological operation is used to remove the noise and close the broken edge lines,and the detected edge lines are filtered and connected.Finally,the first six lines with the same features are selected as Maximum density of stroke features to input.2)This paper proposes a framework of Chinese and Western painting image classification based on stroke features.Through the Chinese and Western painting image database constructed in this paper,the extracted stroke features are input into the convolution neural network combined with SVM classification model.The model shows its unique advantages.Through the experimental comparison with the commonly used ID3,decision tree,KNN,naive Bayes classifier,the classification effect is remarkable.In addition,this paper also compares the classification based on stroke feature with that without stroke feature.The results show that the accuracy of the classification based on stroke feature increases by nearly10%.3)This paper proposes a method to classify western style painting images based on stroke features and color features.Firstly,the western painting image is transformed from RGB model to HSV model,and the K-means clustering method is applied to the generated HSV model.The K value is 20,so as to form 20 kinds of clustering centers.Then the center points of K-means clustering method are sorted from large to small,and the first six colors are selected as the main colors of color features.The experimental results show that the accuracy of stroke classification based on color features is greatly improved.