Objectives  To address significant geometric deformation and radiometric differences between optical and synthetic aperture radar (SAR) images, this paper proposes an automatic registration method based on spatial geometry constraint and structure feature.

  Methods  Firstly, the Harris detector with a block strategy is used to extract evenly distributed feature points in the input images. Subsequently, a local geometric correction is performed for the input image by using rational function model, which aims to achieve local coarse registration between the input image and the reference image. Then, the geometric structural feature descriptor is constructed by using orientated gradient information of images, and the feature descriptor is transformed into the frequency domain, the phase correlation is used as the similarity metric to achieve correspondences by employing a template matching strategy. Finally, the least square method is used to eliminate the mismatches based on spatial geometric constraint relationship between images, followed by a process of geometric correction to achieve the image registration.

  Results and Conclusions  Three sets of high-resolution optical and SAR images were selected as the experimental data. The results the proposed outperform traditional methods in both matching performance and computational efficiency.The proposed method can achieve the registration of optical and SAR effectively and accurately.