Feature guide: a statistically based feature selection scheme

This paper presents a new approach to content-based image retrieval by addressing three primary issues: image feature extraction and representation, similarity measure, and search methods. A statistically based feature selection scheme is introduced to guide the selection of the most appropriate image features for dynamic image indexing and similarity measures. In addition, a fractional discrimination function is proposed to enhance image feature points in conjunction with image decomposition and contextual filtering for image classification. Furthermore, a feature component code is used to facilitate the hierarchical search for the best matching, where images are queried by different features or combinations. The experimental results demonstrate the effectiveness of the proposed method