Authors : Falah H. Abed, Ahmed H. Alboabidallah and Marwah A. Hasan
Abstract: This study investigates a new registration technique for remote sensing satellite images of different spatial resolutions. Existing optimization techniques assume that the accuracy of registration parameters changes by a multivariate normal form. However, when dealing with high-resolution images of highly heterogeneous land cover, the problem of no convergence can occur, this technique overcomes this problem by applying a Monte Carlo simulation. The results show a significant improvement in the correlation coefficients from R = 0.75 to R = 0.92 when a range of satellite images (Landsat_8 and Worldview 3) are used. When the registration results of multi-temporal Landsat_8 images of reduced resolution are compared with the registration of the original resolution images, the spatial accuracy witnessed a substantial improvement of 60% compared with visual registration. Although this technique may be relatively time-consuming compared with previous techniques, the results indicate that the procedure can produce a higher level of accuracy.
Falah H. Abed, Ahmed H. Alboabidallah and Marwah A. Hasan, 2021. Applying of Monte Carlo Simulation to Improved Subpixel Image to Image Registration. Journal of Engineering and Applied Sciences, 16: 258-263.