The field of super-resolution is witnessing significant advancements with a focus on improving the fidelity and consistency of enhanced images and videos. Researchers are exploring innovative approaches to address the challenges of balancing detail recovery and temporal coherence in video super-resolution, as well as developing methods to integrate semantic information and uncertainty awareness in image super-resolution. Notably, the incorporation of diffusion models and 3D consistent representations is leading to high-quality results with improved spatial and temporal coherence. Additionally, efforts to optimize super-resolution models for real-world applications, such as quantization and rotation equivariance, are gaining traction.

Some noteworthy papers in this area include:

• Semantic and Temporal Integration in Latent Diffusion Space for High-Fidelity Video Super-Resolution, which proposes a novel approach to achieve seamless balance between recovering intricate details and ensuring temporal coherence.
• SU-ESRGAN, which introduces a semantic and uncertainty-aware framework for super-resolution of satellite and drone imagery.
• Bridging Diffusion Models and 3D Representations, which presents a 3D consistent super-resolution framework that leverages off-the-shelf diffusion-based 2D super-resolution models.
• QuantVSR, which proposes a low-bit quantization model for real-world video super-resolution.
• Rotation Equivariant Arbitrary-scale Image Super-Resolution, which constructs a rotation equivariant method for arbitrary-scale image super-resolution.