Methoden zur hyperspektralen Bildverarbeitung unter Einbezug von Nachbarschaftsinformation

Hyperspectral imaging and processing allows for non-contact analysis of a plethora of materials and products. Applications can be found for example in food safety, medical imaging and industrial sorting processes. Due to expensive hardware and complex data processing, hyperspectral analysis was mainly limited to a laboratory environment up until a couple of years ago. In the meantime, technical progress has led to real-time inline applications of hyperspectral analysis. This article focuses on methods for hyperspectral image processing and information extraction, namely de noising and spectral unmixing, that incorporate adjacency information. Most existing algorithms ignore this inherent property of hyperspectral images.