Underwater sonar image denoising through nonconvex total variation regularization and generalized Kullback–Leibler fidelity

With the exploration and development of marine resources, sonar imaging technology is gaining increasing attention. Multiplicative speckle noise is often widely distributed in sonar images, significantly degrading their quality. To overcome this challenge, we present a novel model for sonar image denoising, and it consists of nonconvex total variation regularization and generalized Kullback–Leibler fidelity. This model can be efficiently optimized by using the alternating direction method with multipliers based on iteratively reweighted total variation. Moreover, aiming to solve the problems caused by the staircase effect in restored images, the bilateral filter is applied in our method as the post-processor. Experimental results on both simulated and real sonar images demonstrate that the algorithm is effective in eliminating multiplicative speckle noise while maintaining image detail information. To be precise, the subjective visual effect of the denoised images is improved at least 2dB compared with other mainstream methods.