Improving image retrieval using combined features of Hough transform and Zernike moments

In this paper, a novel solution to content based image retrieval system is provided by considering both local and global features of the images. Local features extraction is done by computing histograms of distances from edge lines to the centroid of edge image, where edge lines are detected using Hough transform. It is a robust and effective method to provide association among adjacent edge points, which represent their linear relationship with each other. Zernike moments are used to describe the global features. We have applied algorithms for the fast computation of Hough transform and Zernike moments to make our system fast and efficient. Bray–Curtis similarity measure is applied to compute the similarity among images. A large number of experiments are carried out to evaluate the system performance over six standard databases, which represent various kinds of images. The results reveal that the proposed descriptors and the Bray–Curtis distance measure outperform the existing methods of image retrieval.

► Hough transform (HT) and Zernike moments (ZMs) are used as image features. ► HT works as local feature extractor and ZMs as global one. ► Both HT and ZMs are robust to noise and invariant to geometric transformations. ► Bray–Curtis similarity measure is used to determine similarity between images. ► Retrieval performance is observed to outperform the most recent approaches.