Shatabdi Basu
Department of Computer Science and Engineering, Manipal University Jaipur, Jaipur, Rajasthan, India.

Sunita Singhal
Department of Computer Science and Engineering, Manipal University Jaipur, Jaipur, Rajasthan, India.

Dilbag Singh
Center of Biomedical Imaging, Department of Radiology, New York University Grossman School of Medicine, New York, United States.

DOI https://doi.org/10.33889/IJMEMS.2026.11.3.051

  ;

Abstract

In computational imaging, multi-focus image fusion is a critical process that aims to produce a single image that covers all-in-focus areas from numerous partially focused input images. In this paper, we present a novel approach using a Super-Resolution Generative Adversarial Network (SRGAN) specifically designed for multi-focus image fusion. First, we create a new multi-focus image dataset from the publicly accessible COCO dataset. This process generates a complete collection of annotated image pairs with different focus areas. The generator is designed using a residual learning architecture and upsampling layers. The generator creates a high-resolution fused image with features and texture preservation by processing two input images. Using PatchGAN-based implementation, the discriminator ensures that the fused images maintain global and local coherence through adversarial training. Putting emphasis on intensity, structural similarity, and perceptual qualities, we combine content loss with adversarial loss to achieve balanced learning. Extensive trials on public multi-focus image datasets show that our SRGAN-based model achieves superior fusion quality and texture consistency by outperforming five current state-of-the-art approaches in both quantitative and visual evaluations. The proposed method achieves real-time performance, meets the requirements of contemporary image fusion applications, and demonstrates its efficacy in generating high-quality fused images.

Keywords- Multi-focus image fusion, Deep learning, Generative adversarial network, Super resolution, Perceptual image quality, Computer aided manufacturing.

Citation

Basu, S., Singhal, S., & Singh, D. (2026). MF-SRGAN: A Super-Resolution Generative Adversarial Network for Multi-Focus Image Fusion. International Journal of Mathematical, Engineering and Management Sciences, 11(3), 1227-1264. https://doi.org/10.33889/IJMEMS.2026.11.3.051.