Abstract: A feature extraction and retrieval algorithm based on childblock partition and area division using discrete cosine transformation and singular value decomposition is proposed. First the image is divided into childblocks, discrete cosine transformation is used to extract important coefficients as color feature of the childblock, and then the image is divided into areas, some matrixes are constructed of the discrete cosine transformation coefficients in this area, which are submitted to singular value decomposition to gain the retrieval feature vector to accomplish image retrieval. Experiments show that the proposed algorithm gains good recall and precision, and has a good retrieval effect.

Key words: discrete cosine transformation, singular value decomposition, image retrieval