Fully PolSAR Urban Building Generation via Re-expanding Polarimetric Feature Guided DDPM

The denoising diffusion probabilistic model (DDPM) can produce results of high quality within many fields, such as generating and editing images. Based on the recently proposed method PFG-DDPM, we achieve the urban building generation of fully polarimetric synthetic aperture radar (PolSAR) image in this paper. Briefly, the double-bounce scattering is selected as the guide feature to re-expand the PFG-DDPM model (i.e., proposing a new model rPFG-DDPM with the double-bounce scattering feature Y<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">db</inf> and PFG-DDPM). Besides, to demonstrate its effectiveness, one PolSAR urban building dataset obtained from ALOS PolSAR data is constructed as well. Experimental results indicate that the images generated by our method rPFG-DDPM are much closer to real images than the original DDPM method.