An Energy-Efficient Image Deblurring Accelerator With Quad-Base-Quad-Scale Quantized Format and Layer Normalization-Aware Optimization
- 7 July 2025
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 72 (9) , 1273-1277
Abstract
This brief proposes a novel data-format-based image deblurring accelerator with layer normalization and UNet architecture optimization for mobile cameras. As the demand for photography in dynamic environments continues to grow and the limitations of physical stabilization are tightening, post-processing methods to restore sharp images have gained increasing attention, notably deblurring methods based on convolutional neural networks. However, their heavy computational cost hinders their integration into mobile computing platforms. The proposed accelerator enables energy-efficient acceleration of deblurring through the following three key features: 1) A Quad-base-Quad-scale Quantized format that maintains image quality with only 8-bit, reducing external memory access (EMA) by 33% and achieving 75.7% higher multiply-and-accumulation (MAC) energy efficiency compared to conventional 12-bit precision; 2) A Layer Normalization-Aware Optimization technique, enabling parallel normalization and fusion of affine transformation; 3) A dual-stationary systolic array architecture that selects the optimal dataflow for each UNet block based on processing element (PE) utilization. As a result, the proposed accelerator achieves 2.49 TOPS/W, which is $2.23\times $ higher than prior work, enabling energy-efficient deblurring for mobile applications.Keywords
Funding Information
- Yonsei University Research Fund (2025-22-0142)
This publication has 16 references indexed in Scilit:
- MEGA.mini: A Universal Generative AI Processor with a New Big/Little Core Architecture for NPU2025
- A 28nm 4.96 TOPS/W End-to-End Diffusion Accelerator with Reconfigurable Hyper-Precision and Unified Non-Matrix Processing Engine2024
- SOLE: Hardware-Software Co-design of Softmax and LayerNorm for Efficient Transformer Inference2023
- Simple Baselines for Image Restoration2022
- Uformer: A General U-Shaped Transformer for Image Restoration2022
- Restormer: Efficient Transformer for High-Resolution Image Restoration2022
- NTIRE 2021 Challenge on Image Deblurring2021
- A 3.6 TOPS/W Hybrid FP-FXP Deep Learning Processor with Outlier Compensation for Image-to-Image Application2021
- Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring2017
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015