Fusion of visible and infrared images
In the field of computer vision, convolutional neural networks (CNN) have shown great success due to their capability to extract deep features, which is useful in the fusion of images. Recently there are many existing deep learning fusion methods, however majority of them requires training of t...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Final Year Project |
Language: | English |
Published: |
Nanyang Technological University
2022
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/162845 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | In the field of computer vision, convolutional neural networks (CNN) have shown great success
due to their capability to extract deep features, which is useful in the fusion of images. Recently
there are many existing deep learning fusion methods, however majority of them requires
training of the model, which makes it impractical for real-time use, since it requires a huge
amount of data to train. Furthermore, the fused image often suffers from poor contrast and loss
of fine detail. To address the problem, I proposed a new fusion method which uses pretrained
VGG-19 combined with visual saliency weight map (VSWM) and fast guided filtering (FGF)
that aims to preserves more details and improves the contrast of the fused image. In order to
evaluate the proposed approach, it will be compared against with three other existing fusion
methods based on the quality metrics for images. Finally, we will discuss future work on the
proposed method. |
---|