ABSTRACT

The despeckling of Synthetic Aperture Radar (SAR) images is a critical task. There are certain assessment parameters to judge the algorithm. The article experimentally analyses the dual and diverse nature of Mean Square Error (MSE) in SAR image despeckling. This dual nature is investigated in two different situations. Case 1: when the despeckled SAR image is compared with speckle free SAR image. Case 2: when despeckled SAR image is compared with real speckled SAR image. In both these cases, the comparative statistics of the MSE is different. In case 1, the MSE should be low for better image quality and higher Peak Signal to Noise Ratio (PSNR) and in case 2, MSE must be high for significant speckle reduction. This analysis will help the researchers to use the MSE as performance metric in both these above cases. MSE is a valuable metric in despeckling.