Multi-Focus Image Fusion Based on Multi-Scale Generative Adversarial Network

Area of Science:

• Computer Vision
• Artificial Intelligence
• Image Processing

Background:

• Convolutional neural networks (CNNs) show promise in image fusion due to their information integration capabilities.
• Current CNN-based methods often address only partial aspects of the image fusion process.
• Multi-focus image fusion aims to combine images with different focal planes into a single, sharp image.

Purpose of the Study:

• To propose an end-to-end multi-focus image fusion method using a multi-scale generative adversarial network (MsGAN).
• To leverage multi-scale decomposition combined with CNNs for comprehensive feature utilization in image fusion.

Main Methods:

• Development of a multi-scale generative adversarial network (MsGAN) architecture.
• Integration of multi-scale decomposition techniques with convolutional neural networks.
• Application of an end-to-end learning framework for image fusion.

Main Results:

• The proposed MsGAN method demonstrated superior performance in qualitative and quantitative evaluations.
• Experiments were conducted on both synthetic and real-world (Lytro) datasets.
• The MsGAN achieved better results compared to state-of-the-art multi-focus image fusion techniques.

Conclusions:

• The end-to-end MsGAN is effective for multi-focus image fusion.
• The combination of multi-scale decomposition and CNNs enhances feature utilization.
• The proposed method offers a significant advancement over existing image fusion techniques.