Superpixel linear independent preprocessing for endmember extraction

• DOI: 10.1080/01431161.2023.2274319
• Publication Date: 2023-11-09T10:40:37Z
• Authors: Ricardo Franco; Maria C. Torres-M...; Maria Casamitjana; Tatiana Rondon
• Keywords: Endmember; Multispectral pattern recognition

One of the limitations remote sensing is low spatial resolution open-access multispectral sensors, generating a mixture information. The mixed pixels can be modelled as linear combination fundamental components, called endmember, with weighted contribution or abundance. development unmixing algorithms considering and spectral information has recently increased. Some methods have relied on segmentation to integrate data, one most used superpixel-based segmentation. However, previous work in focuses using superpixels uniform regions. Commonly, hyperspectral imagery, limited literature found images. This paper aims propose new preprocessing approach for Superpixel Linear Independent Preprocessing. proposed generates set candidates endmembers based spatial-spectral information; these are input traditional endmember extraction unmixing. Experimental results show that improves performance extraction.