| As one of the western traditional arts,oil painting has been appreciated and collected by more and more people.By considering the strict standards of oil painting physical preservation,oil painting image digitalization has gradually become the general trend.In digital management,automatic recognition and classification of oil painting art style is one of the key problems to be solved urgently.It is found that most studies generally extract oil painting features without analyzing oil painting artistic style based on art principles,resulting in low validity of features and unable to accurately describe oil painting artistic style.In view of the above problems and considering the artistic principle and process of oil painting,this thesis carries out relevant studies on feature extraction and feature fusion,mainly including:In order to make full use of image information and realize complementary advantages of multiple features,this thesis proposes an algorithm named Adaptive Weight Decision Level Fusion(AWDLF)for oil painting style classification.Firstly,color feature and texture feature are extracted as global features to describe the background color of the image.And LBP feature is extracted as local features to describe the local details of the image.Secondly,the correlation values of the three features are taken as their fusion weights.The naive bayes classifier is used to obtain the posterior probabilities of the categories respectively.And the classification results are fused during the decision-making process.This process makes up for the lack of local details in the existing algorithm and reduces the interference of low-correlation features.Finally,the data set created by FS-Classifier and Pandora18 k data set are used for experiments.Experimental results show that the accuracy of the proposed algorithm is at least 0.48% higher than that of existing algorithms.Aiming at the problem that the existing methods ignore the influence of the main area and the overall effect on the artistic style of the painting,an algorithm named Multi-Feature Fusion Classifier(MFFC)is proposed.Firstly,based on the common arrangement of oil painting elements,the overlapping block method is designed to extract the spatial characteristics of oil painting.This method makes up for the lack of composition style in the existing algorithm,and distinguish the main area and background area at the same time.Secondly,the spatial features and global features are combined in series to increase the location information of elements in the picture.Finally,in the stage of oil painting classification,the spatial voting method is designed.And the classification results of the main area are given priority as the output results of the algorithm,so as to further highlight the role of the main area of oil painting in the classification and realize the automatic recognition and classification of oil painting art style.This algorithm is tested on the data set created by the FS-Classifier model and the public data set WIKIART,and compared with the classification results of other algorithms,the accuracy is improved by at least 13.27%.The feature fusion algorithm proposed in this thesis can achieve accurate classification of oil painting artistic style by fusion and improvement of global features.Experimental results show that the algorithms make full use of image information to realize the complementary advantages of multiple features,and effectively improve the performance of features for oil painting art style classification task,so they have high application value. |
