Context: SAR images can be acquired all time (day and night and whatever the weather) which is a real advantage for tropical areas. Nevertheless the signal is affected by speckle noise which induces strong fluctuations. Nowadays thanks to new constellation, huge multi-emporal stacks are available. This multi-temporal information is of great interest to reduce speckle fluctuations.
Objective: This project is dedicated to multi-temporal denoising of SAR images.
The idea is to improve RABASAR method , which computes the multitemporal mean of the stack and then denoise the ratio image between a noisy data and the stack mean. The filtered mage is then obtained by multiplication of the stack mean and the denoised ratio. It provides satisfying results but some bright residual structures due to the stack mean can appear in the filtered image. In this project the arithmetic mean is replaced by a geometric mean to compute the stack mean. This one is more robust in case of changes in the stack. But the distribution (pdf) of the logarithm of the geometric mean in logarithm is not known and necessary for MuLoG denoiser . The aim of the project is to find a way of computing or approximating this pdf and its derivatives. Comparisons with the original version of the algorithm will be led. The codes of RABASAR and MuLoG are given.
Supervision: Florence Tupin, Nicolas Gasnier
 Ratio-based multi-temporal SAR images denoising,
Weiying Zhao, Loïc Denis, Charles-Alban Deledalle, Henri Maitre, Jean-Marie Nicolas, Florence Tupin
IEEE Transactions on Geoscience and Remote Sensing, 2019
 Charles Deledalle, Loïc Denis, Sonia Tabti, Florence Tupin.
MuLoG, or How to apply Gaussian denoisers to multi-channel SAR speckle reduction?
IEEE Transactions on Image Processing 2017