Logarithmic Mean-Based Thresholding for SAR Image Change Detection

In this letter, a computationally economic and simple logarithmic mean-based thresholding algorithm is proposed for synthetic aperture radar (SAR) image change detection (CD). A difference image (DI) is generated by taking the log ratio of two SAR images. The mean of the logarithm of each image is calculated. Weight is multiplied with the minimum mean to derive weighted threshold. Weight is automatically derived from the mean of the DI. Every pixel in the DI is compared with this derived threshold. Greater magnitude DI pixels are assigned with binary 1 that indicate “changed pixels.” Lesser magnitude pixels are assigned with binary 0 that indicate “unchanged pixels.” Thus, a binary CD map is generated. Experimental results show the effectiveness of the proposed algorithm in terms of false alarms, detection rate, kappa coefficient, and computation time.