S2IP-AS: Asymmetric Synergy with Physics-Inspired Residual Guidance for Defocus Deblurring

Single-image defocus deblurring remains challenging because spatially varying blur affects local structures and global photometric consistency in different ways. Many encoder–decoder restoration networks use the same processing order in the encoder and decoder, which can couple local-structure modulation with color and illumination correction. This paper presents Spatial–Spectral Interaction with Prior-Guided Asymmetric Synergy (S2IP-AS), a framework that organizes these cues with different processing priorities. A Residual Guidance (RG) module estimates residual structural cues from a learnable degraded-feature approximation; its construction is motivated by a local diffusion/Laplacian approximation of defocus attenuation. The encoder then applies intra-scale spectral modulation before prior-guided attention to organize local spectral and phase-sensitive responses, whereas the decoder applies prior-guided attention before spectral modulation to stabilize appearance responses before local refinement. The two paths are linked through skip features and optimized jointly. Experiments report 26.386 dB PSNR on DPDD and 31.079 dB on LFDOF. The LFDOF ordering study favors the proposed assignment in PSNR and LPIPS, while reversed asymmetry has higher SSIM. The resulting design provides a physics-inspired but fully learnable organization of residual, spectral, and photometric cues without imposing a symmetric processing order.

Authors

Institutions

Publication Details

Journal
Electronics
Published
2026-11-19
DOI
https://doi.org/10.3390/electronics15225400
Primary Topic
Advanced Image Processing Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

S2IP-AS: Asymmetric Synergy with Physics-Inspired Residual Guidance for Defocus Deblurring

Kangda Cheng, Haoran Feng, Da Lin, Ruiyang Lin et al.
Electronics
Advanced Image Processing Techniques
article

S2IP-AS: Asymmetric Synergy with Physics-Inspired Residual Guidance for Defocus Deblurring

Kangda Cheng, Haoran Feng, Da Lin, Ruiyang Lin, Weilin Wu, Yitian Zhao, Zhanqi Shi, Miao Zhang
article en

Abstract

Single-image defocus deblurring remains challenging because spatially varying blur affects local structures and global photometric consistency in different ways. Many encoder–decoder restoration networks use the same processing order in the encoder and decoder, which can couple local-structure modulation with color and illumination correction. This paper presents Spatial–Spectral Interaction with Prior-Guided Asymmetric Synergy (S2IP-AS), a framework that organizes these cues with different processing priorities. A Residual Guidance (RG) module estimates residual structural cues from a learnable degraded-feature approximation; its construction is motivated by a local diffusion/Laplacian approximation of defocus attenuation. The encoder then applies intra-scale spectral modulation before prior-guided attention to organize local spectral and phase-sensitive responses, whereas the decoder applies prior-guided attention before spectral modulation to stabilize appearance responses before local refinement. The two paths are linked through skip features and optimized jointly. Experiments report 26.386 dB PSNR on DPDD and 31.079 dB on LFDOF. The LFDOF ordering study favors the proposed assignment in PSNR and LPIPS, while reversed asymmetry has higher SSIM. The resulting design provides a physics-inspired but fully learnable organization of residual, spectral, and photometric cues without imposing a symmetric processing order.

ElectronicsVol. 15(22)
Fujian Normal University (CN), Jimei University (CN)
Advanced Image Processing Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon