Work

Image Deblurring via Frequency-Domain Feature Enhanced Convolutional Neural Networks

Year: 2026

Type: article

Abstract: To address the issues of insufficient restoration of texture details in deblurred images and inadequate learning of frequency domain features, an image deblurring algorithm based on frequency domain feature enhancement and convolutional neural networks is proposed. First, a Fourier residual module with a parallel structure is constructed to achieve collaborative learning and modeling of spatial and frequency domain features. By introducing the Fourier transform, the frequency domain feature learning is enhanced to improve the restoration of texture details. Second, after a gated feed-forward unit is applied to Fourier residual module, its nonlinear representation capability is further improved. In addition, a supervised attention module is introduced at the decoder stage to promote more effective extraction of key features essential for image reconstruction. Experimental results have demonstrated that the proposed algorithm effectively removes blur while better preserving image details.

Source: Sensors

DOI: 10.3390/s26061784

Authors: Yecai Guo, Lixiang Ma, Yangyang Zhang

Institutions: Nanjing University of Information Science and Technology, Anhui Institute of Information Technology, State Key Laboratory of Millimeter Waves

Language: English

Topic: Advanced Image Processing Techniques

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Estimated APC paid: $3,137

Funder: State Key Laboratory of Millimeter Waves