Maximum-A-Posteriori (MAP) estimation Within image processing field, MAP estimation aims to recover inherent clean image Z from observed image G based on a likelihood model and prior knowledge¹: where p(G|Z) defines the probability of a likelihood model and p(Z) is the probability of prior knowledge. The above problem is equivalent to:

InSAR phase restoration based on InSAR-BM3D and DEM generation from filtered phases Background SAR interferometry (InSAR) refers to generating interferograms of SAR images acquired at different times or from changing sensor positions. Geophysical parameters such as elevations and displacement rates over large-scale spatial areas can be retrieved from interferograms, which are crucial…

A lightweight solution for binary segmentation with deep learning Brief review of the challenge https://www.gaofen-challenge.com Dataset RGB optical images from HaiYang-1 satellite Spatial resolution: 50m Images taken over Bohai, China Binary segmentation Number of training samples: 1500 Spatial sizes: 512–2048 Equipment requirements for testing Docker container RAM: 8GB CPU cores: 2 GPU: 1 RTX Titan

An optimization-based SAR time series component decomposition — Advanced satellite and signal processing techniques have made Synthetic Aperture Radar (SAR) capable of monitoring the dynamic processes of the earth with high temporal frequency through co-registered time series. There are several challenges in interpreting these three-dimensional data cubes, including the inherent speckle effect and outliers caused by abrupt changes…

Make a cleaner SAR image based on TV regularization. — Due to the coherent processing of Synthetic Aperture Radar (SAR) image, it is often characterized by strong fluctuations in pixel amplitude, which leads to speckle effect. Within one SAR pixel, there are often plenty of elementary scatterers as illustrated below: Such scattering phenomenon make the phases of these small scatterers…