Blind Video Super-Resolution based on Implicit Kernels

2025-03-10

AI Summary

Existing blind video super-resolution (BVSR) methods often assume spatially invariant blur kernels and neglect spatiotemporal degradation variations, limiting reconstruction fidelity. To address this, we propose a novel framework that explicitly models non-uniform spatiotemporal degradation. Our key contributions are: (1) the first use of implicit neural representations to construct a multi-scale kernel dictionary, enabling dynamic, frame-wise modeling of spatially varying blur kernels; and (2) a recurrent Transformer architecture that jointly performs inter-frame correction and feature alignment via adaptive filtering. This enables more accurate blind degradation estimation and reconstruction optimization under unknown degradation conditions. Extensive experiments demonstrate state-of-the-art performance on three benchmark datasets—REDS, Vimeo-90K, and BSDS—outperforming prior methods including FMA-Net, with up to 0.59 dB PSNR gain. The source code is publicly available.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Blind video super-resolution (BVSR) is a low-level vision task which aims to generate high-resolution videos from low-resolution counterparts in unknown degradation scenarios. Existing approaches typically predict blur kernels that are spatially invariant in each video frame or even the entire video. These methods do not consider potential spatio-temporal varying degradations in videos, resulting in suboptimal BVSR performance. In this context, we propose a novel BVSR model based on Implicit Kernels, BVSR-IK, which constructs a multi-scale kernel dictionary parameterized by implicit neural representations. It also employs a newly designed recurrent Transformer to predict the coefficient weights for accurate filtering in both frame correction and feature alignment. Experimental results have demonstrated the effectiveness of the proposed BVSR-IK, when compared with four state-of-the-art BVSR models on three commonly used datasets, with BVSR-IK outperforming the second best approach, FMA-Net, by up to 0.59 dB in PSNR. Source code will be available at https://github.com.

Problem

• Enhance low-resolution videos in unknown degradation scenarios.
• Address spatio-temporal varying degradations in video frames.
• Improve performance over existing blind video super-resolution methods.

Innovation

• Implicit neural representations for kernel dictionary
• Recurrent Transformer for coefficient prediction
• Multi-scale kernel dictionary for BVSR