Region based multi-spectral fusion method for remote sensing images using differential search algorithm and IHS transform

• Published in: Expert systems with applications 2022-03, Vol.189, p.116135, Article 116135
• Main Author: Kurban, Tuba
• Format: Article
• Language: English
• Subjects: Clustering; Computer vision; Differential search algorithm; Feature extraction; Image classification; Image fusion; Image processing; Image resolution; Image segmentation; Landsat satellites; Multi-spectral imaging; Optimization; Remote sensing; Satellite imagery; Saturation (color); Search algorithms; Spectra; Spectral bands
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2021.116135

•A two-step cascade method to fuse 5 different bands of Landsat 8 images.•Adaptive region segmentation and optimal weighting for each region.•Differential Search algorithm (DSA) used to maximize the fusion quality.•RGB bands are combined with Intensity-Hue-Saturation (IHS) transform.•Superior visual performance compared with 11 methods with 5 objective metrics. Remote sensing applications such as clustering, classification, feature extraction, measurement and change detection need fused high-resolution images that contain complementary information coming from multi-spectral bands. To transfer as much information as possible from source images to the fused image, image fusion can be considered as an optimization problem. In this paper, a new, two-stage, multi-spectral, region based, optimal fusion method is proposed. Panchromatic and infrared bands of Landsat 8 satellite images are fused with K-medoids segmentation and Differential Search Algorithm (DSA). The fused image is combined with the red, green, and blue (RGB) bands of the same area with intensity-hue-saturation (IHS) transform. Experiments carried out on test images indicate that developed method performed better than most of the state-of-the-art fusion techniques in terms of both visual and numerical evaluations.