A novel approach for despeckling of sonar image

Abstract

Speckle noise in the sonar images causes incorrect interpretation and analysis. It is necessary to extract all essential information from the affected image. From the literature studies carried out, it is found that there are various techniques available for the removal of speckle noise in sonar images. In this paper, the removal of speckle noise in sonar images using spatial filters such as Lee, Frost, Median, Wiener, Mean, Speckle Reducing Anisotropic Diffusion (SRAD) and in frequency domain using Discrete Wavelet Transform (DWT)-Haar along with median, and SRAD has been done. A hybrid technique has been introduced in spatial domain and then a comparative study has been done for all the filters. The performance has been analyzed with the help of quality metrics such as Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), Visual Information Fidelity Criteria (VIF), Structural Similarity Index Metrics (SSIM) and Correlation. The result concluded that the hybrid of SRAD and Gaussian Filter in spatial domain proves best for the filtering of speckle noise in sonar images among all filters tested here.