A Comparative Assessment of Selected Algorithms on Multi Domain Registration of Remote Sensing Images

With the development of remotely sensed data acquisition techniques, the integration of complementary data has found a key role in many applications. The prerequisite of the data integration is the data registration. There are many approaches for image registration that they can be categorized based on the nature of data sources as the multi domain and single domain methods. The multi domain methods employ the data sources with different properties of gray values such as irradiance, the return strength of the laser pulse, and the height values. The different aspects of multi domain registration methods include the matching algorithm as feature-based or intensity-based, the level of automation, availability of initial registration parameters, the implementation cost, and so on. In this study, a number of multi domain registration algorithms are selected and compared based on the mentioned aspects and then, an automatic multi domain registration method is proposed to register an intensity image of LiDAR and a satellite image without using initial registration parameters. For this, the combination of the Scale Invariant Feature Transform (SIFT) detector and the mutual information theory is employed to use the strengths of both feature-based and intensity-based matching algorithms to decrease the difficulty of multi domain image registration. The final registration results prove that the combination of the feature-based and the intensity-based matching algorithms could be an efficient solution of multi domain image registration problem, especially for registration between the LiDAR intensity and the satellite images with the RMSE of one pixel.