Change detection in remote sensing image using a modified logarithmic mean-based thresholding

In this paper, we propose a novel approach to change detection in remote sensing imagery by modifying the logarithmic mean-based thresholding technique (MLMBTICD). This method introduces a preprocessing step using a mean filter to enhance the accuracy of detecting changes between multi-temporal satellite images. The mean filter reduces noise and smoothens the images before calculating the logarithmic difference, which improves the quality of the change detection process. The proposed approach was tested on two benchmark datasets: the Onera dataset, which contains satellite images of urban regions, and the Ottawa dataset, consisting of RADARSAT-2 images. The effectiveness of the MLMBTICD method was evaluated using Overall Accuracy (OA) and Kappa metrics. The results demonstrate that our method achieves better performance compared to the original logarithmic thresholding method, yielding improved change detection accuracy. The preprocessing step significantly enhances the quality of the detected changes, making the proposed method a robust and efficient solution for various remote sensing applications, including land use monitoring, urban development, and environmental change analysis.