Decorrelation stretch method

I am trying to replicate some work by Rajendran, S. et al, to apply it to the use of Sentinel-1 and Sentinel-2 combined to track oil with a higher temporal resolution than using both independently.

However I am struggling with what he means by ‘decorrelation stretch’. I am also not getting anywhere near the same visual results that they achieve in their paper.

I have set the band ratios as per the paper, but I seem to be getting stuck at the ‘decorrelation stretch method’.

Here is the link: Sentinel-2 image transformation methods for mapping oil spill – A case study with Wakashio oil spill in the Indian Ocean, off Mauritius - ScienceDirect

Any help would be appreciated!