Image Retrieval Based on NSCT from the Science and Technology Resource Image Database

Considering the characteristics of science and technology resources image, the system structure of image retrieval techniques from the science and technology resources image database is designed to rapidly and effectively retrieve the image from the database. Firstly, a novel image retrieval algorithm based on texture is proposed. The algorithm proposed takes fully the advantage of statistical attribution of image's nonsubsampled contourlet transform (NSCT) coefficients to form the eigenvector for depicting textural feature and shape feature. Finally, the TBIR method is used to retrieve the image from the database. Secondly the CBIR method is used to retrieve the image from the result of the TBIR method. Moreover, the three key technologies, the image segmentation, the feature extraction and the similarity measurement, is researched and comprehensively used to retrieve the image.