A NOVEL SOCIAL IMAGE DATA MINING ALGORITHM BASED ON MULTI-MODEL DATA FUSION

A multi-model data fusion algorithm based on social image features and its corresponding text description is proposed. Firstly, three different features are investigated, and particularly, uniform colour space (UCS) is used to improve performance of colour pattern recognition. Similarity of different images is measured using the text associated with social image information, which can enhance the distinguish ability of the original image features. The experimental results show that compared with using text feature or image features, combining two kinds of information can achieve better clustering performance. The proposed clustering algorithm based on pre-classification and multi-model data fusion is better than commonly used clustering algorithms at present.