Image Annotation Based On Ensemble Of Naive Bayes Classifier

With the rapid development of the internet technology and data storage technology, more and more information is shown in the form of images, so it is very necessary to realize a more effective retrieval in large-scale digital image. Semantic clarity is an important prerequisite of the large-scale digital image management, but the existing research shows that there is a significant gap between the content based image retrieval and advanced semantic interpretation of images by humans, which also means that it exists the semantic gap between the underlying features of the image and the advanced semantics of the image understood by human. The technology of automatic image annotation automatically obtains the semantic key words of images from visual characters, which is very important to realize automatic image annotation through the machine learning method.The target of research is to realize image annotation using the machine learning algorithm. The research has selected the naive bayes algorithm and realized automatic image annotation through completing the design and implementation of the naive bayes classifier. Automatic image annotation based on the classifier ensemble method, which obtains high-level semantic keywords through processing and analyzing the underlying visual information features of image. In this paper, we propose an ensemble construction method of the naive bayes classifier, which concentrates on how to build diverse classifier and reconstructs a full feature set for each classifier in the ensemble. The model of automatic image annotation can be obtained through training the ensemble classifier. In this paper, the automatic image annotation based on the ensemble classifier, which realizes that image annotation is transformed to classification of the underlying visual information features of image. In this paper, we design a Web image retrieval system platform, which allows images to be searched in the same way as we search text documents. In conclusion, this system brings a great convenience to the future research work. Meanwhile, it is a learning to realize the image retrieval of commercial application.In this paper, experiments are performed to verify the effectiveness of the algorithm with a standard UCI dataset and a standard Corel dataset. The experimental results show that the proposed algorithm can improve the classification accuracy very well. At the same time, automatic image annotation system based on the ensemble classifier perfectly achieves the automatic image annotation.