A diffusion model that generates latent high-frequency maps from low-quality stereo images and injects them into a transformer restoration network yields modest gains on stereo super-resolution, deblurring, and low-light enhancement.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
DiffStereo: High-Frequency Aware Diffusion Model for Stereo Image Restoration
A diffusion model that generates latent high-frequency maps from low-quality stereo images and injects them into a transformer restoration network yields modest gains on stereo super-resolution, deblurring, and low-light enhancement.