Recognition and Analysis of Painting Styles with the Help of Computer Vision Techniques

With the continuous improvement of living standards, it has become a trend to enhance one's cultural literacy. However, there is a pressing need to find a solution to the issue of how to comprehend the artistic features of the pieces. Regard of this, the study uses computer vision technology to identify and analyze the paintings' style and mood.. First, a painting style classification recognition model based on multi-feature fusion is built by using wavelet transform for denoising and fusing color, texture and spatial features of paintings. Then, the dataset is expanded using style migration, and style features and high-level semantic features are collected to build a painting sentiment analysis model based on style migration. The results indicated that the classification accuracy and recall of the proposed painting style classification and recognition model were consistently above 95%. On the WikiArt dataset and the OilPainting dataset, the subject manipulation characteristic curves of the proposed model were able to completely cover the results of the other three models with the best overall performance. The proposed painting sentiment analysis model demonstrated good recognition performance on both datasets, with recognition accuracy above 90%. The results of the study help to accurately categorize the styles and recognize the emotions of paintings, providing certain convenience for the appreciation of art works.