My project proposal concerns the improvement and the implementation of new functionalities for the OSSIM plug-in that I developed in the last two years in the framework of the 2014 and 2015 GSoC programs.
I was thinking, first of all, to finalize and consolidate all the work done so far, specifically, to:
- complete the workflow for generating DSMs from SAR images
- improve the automatization of the pyramidal approach implemented for DSM generation
The main aim would then be to implement a processing tool able to merge various DSMs resulting from different pairs or triplets, acquired with different sensors or in different period in times. This is particular useful for SAR images for which usually you've got a stack of images over the same area.
In particular, the idea is to create a new function able to generate a mask for exploiting only the best areas in the DSMs merging phase. Specifically, in SAR images lay-over and fore-shortening areas would be automatically detected in the input images; a mask can hence be generated and applied to the corresponding DSM in order to cross out those pixels not "good" and replace them with pixels more reliable coming from other DSMs.