Abstract: The invariant of Radon transform using the statistical theory is constructed and a new kind of color image retrieval algorithm based on Radon transform is presented. In the first place, the shift invariant project is achieved by centering the Radon project of HSV color space. The singular values of the project can be used as the character vector. The singular values of the scaling invariant can be obtained by the statistical characteristics of Radon transform. Since the singular values of the matrix are uncorrelated with the position of the column or the row of the matrix. In this way, the character vector of singular values with shift, scaling and rotational invariant is arrived at by this algorithm. The Gaussian model is used to normalize the different sub-character distances to the character vector. The similarity between the querying image and other images is computed by the Euclidean distance. Experiments indicate that this method has a higher retrieval-rate than other methods.

Key words: content-based image retrieval, radon transform, singular value decomposition