Investigation of TsGAN-based multimodal image fusion to augment image pre-processing abilities
Abstract Image processing techniques are broadly used across various fields; however, their effectiveness is recurrently constrained by the limitations associated with images from single source. Image fusion offers robust solution by integrating the data from multiple sources to produce more compreh...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
SpringerOpen
2025-07-01
|
| Series: | Journal of Electrical Systems and Information Technology |
| Online Access: | https://doi.org/10.1186/s43067-025-00229-6 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Image processing techniques are broadly used across various fields; however, their effectiveness is recurrently constrained by the limitations associated with images from single source. Image fusion offers robust solution by integrating the data from multiple sources to produce more comprehensive and informative output. Diverse types of images can be used for the purpose, such as RGB images, LiDAR images and some medical imaging modalities such as CT scan, PET scan, MRI scan. Use of diverse modality images highlights the essential limitations of single source imagery and highlights the critical role of fusion techniques in enhancing data quality and enabling more effective information extraction. An extensive range of image fusion techniques is available, across from traditional approaches such as intensity-hue-saturation (IHS), principal component analysis (PCA), discrete cosine transform (DCT), and discrete wavelet transform (DWT) to more advanced methods such as deep learning (DL) and generative adversarial networks (GANs). While DL has shown promise in image fusion, it is often hindered by significant computational demands and large data requirements. In contrast, GANs, with their dual network architecture and adversarial training process, present a promising alternative with the potential for superior fusion performance. The study proposes a novel method for fusing visible and infrared images using TsGAN, specifically developed for image pre-processing. The TSGAN framework generates a unified texture map that captures essential gradient preservation into the generator’s loss function, the model successfully retains critical information about texture from the input source images. Additionally, “a multiple decision map-based strategy” is introduces for fusion to enhance texture extraction. Empirical evaluations confirm the effectiveness of the proposed approach, highlighting its superiority over existing algorithms in both qualitative and quantitative analysis. This method is particularly significant in image processing applications where texture information is critical. Additionally, paper suggests expanding the application of these techniques to areas such as precision agriculture, object detection and surveillance, where the extraction of texture-based information is crucial. |
|---|---|
| ISSN: | 2314-7172 |