Research On Chinese Painting Feature Extraction And Classification Recognition Algorithm

Chinese painting is the quintessence of the nation and crystallization of ancient Chinese painting art.With the development of the times,Chinese painting has become a kind of unique Chinese painting art.At present,most researches on classification and identification of traditional Chinese paintings are based on the texture features of Chinese painting images.First,the original Chinese painting images are converted to grayscale images,and then extract the texture feature of images.However,certain information included in images will be lost during the process of the grayscale conversion.This will affect the accuracy of feature extraction and classification recognition.There are few researches on the extraction of Chinese painting's color features.The existing color feature extraction algorithms are basically based on a sole color space,and often ignoring the spatial characteristics of the image.The main research of this paper focuses on the extraction of texture features and color features of Chinese painting.The main contributions of this paper are as follows.1)A method for extracting texture features is proposed in this paper.The method omits the gradation conversion of the images and directly against original Chinese painting images.This approach maximally reserved the information contained in the original image.Using the multi-color analysis and multi-scale analysis the algorithm combines traditional gray level co-occurrence matrix to extract texture features.Experiments show that the multi-scale gray level co-occurrence matrix algorithm is superior to the traditional gray level co-occurrence matrix algorithm and color gray level co-occurrence matrix algorithm.2)A mixed color space is proposed in this paper.Three color spaces which are close to the human visual perception system: HIS,HSV,and YUV are selected.Combining the respective advantages of these color spaces constitutes the mixed color space.This mixed color space makes up the inadequacies of a sole color space.A new block method is proposed to solve the lack of image space information in the feature extraction process of color moments.Finally,the algorithm is validated by combined the gray-level co-occurrence matrix algorithm.Experiments show that the mixed color space block color feature extraction algorithm is superior to the sole color space color feature extraction algorithm.3)The classification requirements for the Chinese painting database are generally divided into two categories,which are classified according to subject matter and classified by author.In this paper,we carried out relevant experiments for these two classification requirements.We used the integrated feature vectors of fused textures and color features to classify and recognize the Chinese painting images,which showed the validity of the two algorithms proposed in this paper.