Landmark Guided HINet for Facial Image Deblurring

DOI

10.2991/978-94-6239-693-7_76How to use a DOI?

Keywords

Normalisation; HINet; Facial Landmarks; CNN; Image De-blurring

Abstract

Facial images captured in real life often get blurred due to motion blur or poor focus. Facial image deblurring plays an important role in surveillance, authentication and biometric applications. Although conventional Convolutional Neural Network (CNN) and Generative Adversarial Network (GAN) based deblurring methods achieve promising results, they fail to preserve facial geometry because normalization dis- torts important facial details. To resolve this limitation, a lightweight data-driven Landmark Guided Half Instance Normalization (HIN) architecture is proposed, which enhances the Half Instance Normalization Network (HINet) deblurring model by adding facial landmarks as structural guidance. In the proposed approach, facial landmarks act as guidance that inform the network which regions contain critical facial de- tails. HIN is applied to non-facial areas, which then helps the model to preserve important features like eyes, nose and mouth. Also, land- mark prediction branch is added with a heatmap-based landmark loss. Experimental results show that proposed model produces sharper and structurally consistent face restorations. Quantitative experiments show that the model achieves a Peak Signal-to-Noise Ratio (PSNR) of 26.2 and Structural Similarity Index Measure (SSIM) of 0.91, which indicates improvements. These results demonstrate that integrating facial priors improves recognition accuracy, supporting intelligent and sustain- able surveillance systems.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Rutuja Dautpure
AU  - Sakshi Powar
AU  - Sayali Khedkar
AU  - Vaishnavi Shelar
AU  - Amit Joshi
AU  - Soma Ghosh
PY  - 2026
DA  - 2026/06/16
TI  - Landmark Guided HINet for Facial Image Deblurring
BT  - Proceedings of the International Conference on Intelligent Systems for a Sustainable Future (ISSF 2026)
PB  - Atlantis Press
SP  - 769
EP  - 784
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-693-7_76
DO  - 10.2991/978-94-6239-693-7_76
ID  - Dautpure2026
ER  -