due to the side-looking geometry of SAR systems, the are acquired in slant range geometry. This means, some areas are farer away from the sensor than others. This consequently distorts the image and leads to different pixel sizes. By applying a terrain (or ellipsoid) correction this can be changed so the image ends up in squared pixels and without distortions (ground range geometry).But that step involves resampling the pixels, meaning that their values are slightly changed due to geometric and radiometric corrections.

Most steps in SAR remote sensing, such as filtering or calibrating, were initially designed to be performed on slant-range images. So it is quite common in SAR pre-processing to leave the image in slant geometry as long as possibe and terrain correct and project it to a coordinate system as a last step. This guarantees that you work on the real pixel values instead of already altered radiometry.

Of course, in the image metadata there is information on the geocoding of the images that can be used for the geometric projection. But at first, SAR data is slant-range. Many images can be ordered as ground range products as well, but that means that they were already processed (to an often not known degree).

Depends on what you need the data for.

Please have a look at this topic:

http://forum.step.esa.int/t/gtc-geocoded-terrain-corrected-data/3472/2

Thank you very much sir #Andreas Braun. Very useful information.

Sir could you please elaborate any Image Registration Application ? This helps me a lot.

if you need several SAR images perfectly aligned co-registration is the right choice. It targets smaller shifts caused by differences in incidece angle and other factors and kind of distorts all slave image according to the geometry of a chosen master image.

Here is a quite interesting tutorial on that using the free ESA SNAP software.

https://sa.catapult.org.uk/wp-content/uploads/2016/05/SNAP-Sentinel-1_TrainingCourse_Exercise2.pdf