A Painting Image Retrieval Approach Based on Visual Features and Semantic Classification

Traditional content-based image retrieval technology does not support semantic-based image retrieval, and the target image based on visual features often lacks the understanding of its image semantics. This article mainly studies the use of visual and semantic features of images synthetically and realize the classification and retrieval of image database based on semantic classification of image database. In the visual feature extraction of image, three visual features are introduced. Then, the learning method of support vector machine(SVM) is applied to image classification, and the visual feature retrieval results are filtered by the classification results of image database. A multi-class classifier based on SVM is designed, which uses semantic and visual features to represent image content, trains and learns SVM, and classifies image database semantically by the image classification model. Finally, an image retrieval system is designed and the experiments on color, texture and comprehensive features are performed. The results show that the classification retrieval approach is effective and the retrieval accuracy is improved.