DiNAT-IR: Exploring Dilated Neighborhood Attention for High-Quality Image Restoration

TL;DR

DiNAT-IR introduces a novel dilated neighborhood attention mechanism combined with a channel-aware module, enhancing global and local feature integration for improved high-quality image restoration.

Contribution

This paper proposes DiNAT-IR, a Transformer-based model that effectively balances global context and local details using dilated neighborhood attention and a channel-aware module for image restoration.

Findings

Achieves competitive results on multiple image restoration benchmarks.

Effectively balances global context and local detail in high-resolution images.

Demonstrates the effectiveness of dilated neighborhood attention in low-level vision tasks.

Abstract

Transformers, with their self-attention mechanisms for modeling long-range dependencies, have become a dominant paradigm in image restoration tasks. However, the high computational cost of self-attention limits scalability to high-resolution images, making efficiency-quality trade-offs a key research focus. To address this, Restormer employs channel-wise self-attention, which computes attention across channels instead of spatial dimensions. While effective, this approach may overlook localized artifacts that are crucial for high-quality image restoration. To bridge this gap, we explore Dilated Neighborhood Attention (DiNA) as a promising alternative, inspired by its success in high-level vision tasks. DiNA balances global context and local precision by integrating sliding-window attention with mixed dilation factors, effectively expanding the receptive field without excessive overhead. However, our preliminary experiments indicate that directly applying this global-local design to the classic deblurring task hinders accurate visual restoration, primarily due to the constrained global context understanding within local attention. To address this, we introduce a channel-aware module that complements local attention, effectively integrating global context without sacrificing pixel-level precision. The proposed DiNAT-IR, a Transformer-based architecture specifically designed for image restoration, achieves competitive results across multiple benchmarks, offering a high-quality solution for diverse low-level computer vision problems.