Statistical Latent Manifold-Guided Framework for Generative Super-Resolution

Abstract

We present a statistical latent manifold-guided framework for generative super-resolution, designed to improve reconstruction quality and structural consistency under challenging low-resolution conditions. The method incorporates manifold-aware latent constraints to preserve clinically meaningful details while reducing hallucination artifacts. The framework was presented at KCC 2025 poster session by Seo-Yeon Choi as first author, with Kyungsu Lee as corresponding author.