Implicit Neural Representation for Video and Image Super-Resolution

2025-03-06

AI Summary

Existing video and image super-resolution methods suffer from poor detail consistency, temporal instability, and reliance on computationally expensive optical flow estimation. To address these issues, this paper proposes SR-INR—the first unified implicit neural representation (INR) framework for both image and video super-resolution. SR-INR conditions a lightweight, end-to-end differentiable MLP solely on low-resolution inputs and 3D spatiotemporal coordinates, directly regressing high-resolution pixel values without explicit motion modeling. Its core innovation is a joint spatiotemporal coordinate encoding scheme, which ensures inter-frame detail consistency and strong temporal stability while preserving architectural simplicity. Experiments demonstrate that SR-INR achieves or surpasses state-of-the-art performance across multiple benchmarks. Notably, it attains superior computational efficiency—featuring significantly fewer parameters, faster inference, and reduced computational overhead—while maintaining high-fidelity spatiotemporal coherence.

Technology Category

• Computer Vision: Representation Learning for Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Machine Learning: Representation Learning

Application Category

• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

We present a novel approach for super-resolution that utilizes implicit neural representation (INR) to effectively reconstruct and enhance low-resolution videos and images. By leveraging the capacity of neural networks to implicitly encode spatial and temporal features, our method facilitates high-resolution reconstruction using only low-resolution inputs and a 3D high-resolution grid. This results in an efficient solution for both image and video super-resolution. Our proposed method, SR-INR, maintains consistent details across frames and images, achieving impressive temporal stability without relying on the computationally intensive optical flow or motion estimation typically used in other video super-resolution techniques. The simplicity of our approach contrasts with the complexity of many existing methods, making it both effective and efficient. Experimental evaluations show that SR-INR delivers results on par with or superior to state-of-the-art super-resolution methods, while maintaining a more straightforward structure and reduced computational demands. These findings highlight the potential of implicit neural representations as a powerful tool for reconstructing high-quality, temporally consistent video and image signals from low-resolution data.

Problem

• Enhances low-resolution videos and images using implicit neural representation.
• Achieves high-resolution reconstruction without optical flow or motion estimation.
• Provides efficient, temporally stable super-resolution with reduced computational demands.

Innovation

• Uses implicit neural representation for super-resolution
• Leverages 3D grid for high-resolution reconstruction
• Achieves temporal stability without optical flow