Deep learning-based compression and despeckling of SAR images

Inhalt:
Combining despeckling and compression tasks is worthwhile because a decrease in the amount of information to be encoded will result in a more efficient data downlink. This paper presents a self-supervised solution to performing joint compression and despeckling of SAR images, with an estimation of the reflectivity based on an original adaptation of recent machine learning-based advances in the fields of image compression and SAR images despeckling. The proposed solution was successfully tested on real-world data from TerraSAR-X, showing great potential for achieving state-of-theart despeckling under the constraints of end-to-end optimized compression with variational autoencoders.