Remote Sensing Image Fusion Based on Morphological Convolutional Neural Networks with Information Entropy for Optimal...

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Remote sensing image fusion is crucial for extracting detailed information from multispectral and panchromatic images.
• Existing fusion methods often struggle to balance spatial detail preservation with spectral fidelity.

Purpose of the Study:

• To develop an advanced remote sensing image fusion method utilizing optimal scale morphological convolutional neural networks (CNNs).
• To leverage information theory and entropy principles for enhanced image component decomposition and fusion.

Main Methods:

• Employing sparse decomposition-morphological component analysis (MCA) with dual dictionaries (DCT and curvelet transforms).
• Determining optimal decomposition scale and threshold via information entropy maximization.
• Utilizing an attentional CNN to fuse extracted cartoon and texture components.

Main Results:

• The proposed method effectively retains original image information.
• Significant improvements in spatial resolution and spectral fidelity were observed.
• The fusion process successfully integrates optimal cartoon and texture components.

Conclusions:

• The optimal scale morphological CNN-based fusion method offers a new paradigm for remote sensing image fusion.
• This approach enhances image quality by preserving spectral information and improving spatial resolution.
• The study highlights the potential of multi-morphological deep learning in image fusion applications.