Effects of compression on SAR image interpretability

SAR images relevance for IMINT (IMagery INTelligence) and civil protection applications has increased in the last years, especially regarding the provision of information to decision makers acting in Area of Operations outside the national borders. Therefore, it is necessary to send information, as well as images, to the Area of Operations through Telecommunication satellites. The large data amount of new systems requires both large bandwidth allocation and transmission time, clearly unacceptable for near real-time applications. Image compression could be a suitable solution, but strict image-interpretability limitations request the saving of critical details. This research analyzes the effect of JPEG and wavelet-based compression on interpretability of SAR images provided by COSMO-SkyMed (CSK) system and Sandia National Laboratories. Images were given to a set of military analysts who expressed a subjective judgement on the image quality and used them to carry out actual detection tasks. Both subjective and objective results show that a moderate amount of compression is perfectly acceptable and does not entail measurable losses in the ability to detect targets of interest. The wavelet-based techniques provide superior performance w.r.t. JPEG, as expected, and the use of ROIs further improves the experimental results.