Blind Remote Sensing Images Quality Estimation and Denoising

The Remote Sensing Images are always interfered by various noises for the complex environment of atmosphere. Despite the success of both residual and dense skip connections in image denoising, both aggregation types have drawbacks. In this work, we propose a variational posterior evaluation and blind image denoising (VPE-BID) algorithm. By using the improved Bayesian framework based on the variational inference method and the multi-scale structure similarity (MS-SSIM) loss function, we take the inherent clean image and noise variance as the potential variables of the input noise image to deparameterize a posteriori. This posterior provides explicit parametric forms for all its involved hyper-parameters, and thus can be easily implemented for blind image denoising with automatic noise estimation for the test noisy image. Extensive experiment on UCMerced LandUse database and BSD68 demonstrates the proposed algorithm has the most advanced performance in qualitative and quantitative measurement.