Fusion of Synthetic Aperture Radar Data and Optic Images based on Curvelet Transform

Satellite remote sensing (RS), gathered data with different spatial and spectral characteristics of objects or phenomena from a distance that each of them represents part of the object properties. Although multispectral data gives rich spectral information from objects, but significantly influenced by environmental factors such as smoke, fog, clouds and the sunlight. In contrast to optical sensor, the virtual aperture radar sensors have the ability to take data in all types of weather conditions or day and night. Synthetic Aperture Radar (SAR) data can highlight the structural and textural details in the image. It is sensitive to terrain components of shape, direction, roughness, and moisture. So optical data provide detailed spectral information useful for discriminating between surface cover types, while the radar imagery highlights the structural detail in the image. Therefore image fusion techniques can help us for combining of different properties of optical images and SAR data that it can give us a complete view of the target and present higher accuracy and reliability of results obtained by this method. Curvelet transformation is more suitable in comparison with many other transformations for analysis of curved edges, high precision to approximate, describe scattering and directions. In this paper, transition SAR and optical images to Curvelet space by using Curvelet transformation, then the weighted average method is used for fusion in Curvelet space and finally fused images obtained by applying a reverse Curvelet transformation. Our case study is Shiraz city that we used data from this city for implementation of proposed method. Statistical methods and classification were used to evaluate the fused images. IHS and Wavelet transform methods is used for comparison to proposed method. Statistical parameters include standard deviation, entropy, standard spatial frequency, correlation and image quality index show improvement of fused images by proposed method than other methods. Considering that accuracy of classification depends on the image spatial and spectral information, for evaluate the effectiveness of fusion on the spatial and spectral resolution, images are classified. With the classification of input optics image and fusion image, overall accuracy improved 4 percent and Kappa coefficient increased 0.05 compared to the input image. The results show the suitability of the proposed algorithm for fusion SAR and optical images.