Enhancing Vein Detection in Hyperspectral Images Using Correlation-PCA Fusion and Clustering

Hyperspectral imaging plays a pivotal role in various fields, particularly in precise human vein detection within medical diagnostics. However, dealing with the large-scale hyperspectral (HS) data presents challenges. To address this, dimensionality reduction techniques are commonly employed to enhance data manageability during processing. The introduced novel dimensionality reduction approach, Correlation-PCA Fusion and Clustering (CoPCA-Clus), is rigorously compared with established techniques, namely Principal Component Analysis (PCA) and Folded PCA (FPCA), specifically for vein detection in HS images. Results demonstrate that CoPCA-Clus surpasses PCA and FPCA, exhibiting superior performance across accuracy, precision, recall, false positive rate, and false negative rate. Additionally, performance metrics are derived for each technique, and classification images are generated. Subsequently, morphological operations enhance the visualization of vein regions within the HS image.