A Remote Sensing Image Retrieval Method Based on Quaternion Transformation

Remote sensing image retrieval requires that image features have the properties of rotation and scale invariance. Currently different color channels are often processed separately during the process of feature extraction, and unique geometric features of remote sensing image are utilized inadequately, which decrease image retrieval precision. A remote sensing image retrieval method based on quaternion transformation is proposed. Aiming at solving the problem of texture features extracted based on quaternion orthogonal Fourier-Mellin moments only have rotation invariance, we construct texture feature with rotation and scale invariance based on quaternion orthogonal Fourier-Mellin moments. We also use quaternion orthogonal Fourier-Mellin moment to detect image edges, get edge color image and extract edge color histogram, and integrate different features for image retrieval. The experimental results show that the proposed method has good robustness to image rotation and scale, and has better retrieval performance than other methods.