Image retrieval method based on fast equal-average K nearest neighbor search and query-feature-vector-recomposition feedback

For improving the efficiency of the content-based image retrieval (CM), a novel fast retrieval algorithm named equal-average k nearest neighbor image retrieval (EKNNIR) is proposed in this paper This algorithm can effectively decrease the computational complexity, while not reducing the accuracy of retrieving the K nearest similar images. In order to control the proportion of the components of the feature vector, Gaussian normalization is adopted. Furthermore, a simple query-feature- vector-recomposition feedback method is proposed to improve the retrieval precision. Experimental results show that the proposed method can speed up the retrieval process and achieve satisfying retrieval precision, especially for large image database and high feature vector dimension.