Please use this identifier to cite or link to this item: http://dspace.dtu.ac.in:8080/jspui/handle/repository/21663
Title: SPECTRAL UNMIXING USING MODIFIED PICTURE FUZZY CLUSTERING METHOD
Authors: JAIN, VATSAL
Keywords: SPECTRAL UNMIXING
(PFCM) PICTURE FUZZY C-MEANS
IRCM METRICS
BVI
Issue Date: May-2025
Series/Report no.: TD-7864;
Abstract: Unsupervised clustering methods are used in spectral unmixing of multispectral images since to estimate the endmembers. But their performance is hampered due to inadequate data and imprecise initialization of endmembers. In this study we propose modified picture fuzzy c-means (PFCM) technique to initialize the endmembers. Experimental assessments carried out on multi-spectral datasets reveal that our approach consistently surpasses current leading techniques. Performance indicators such as the BVI, UI, and IRCM validate its enhanced capabilities in endmember separability, accuracy of abundance estimation, and spatial consistency. These findings underscore the efficacy and adaptability of the proposed method across various multi-spectral imaging contexts. Incorporating the heuristic consistently enhances performance in various datasets. Significantly, a revised version of PFCM that includes our heuristic achieving a BVI of 0.004, UI of 0.999 and IRCM of 0.466 on the multispectral image dataset. Tests performed on both the Iris dataset and a multispectral image reveal enhancements in the BVI, UI, and IRCM metrics, as well as faster convergence rates compared to traditional implementations.
URI: http://dspace.dtu.ac.in:8080/jspui/handle/repository/21663
Appears in Collections:M Sc Applied Maths

Files in This Item:
File Description SizeFormat 
vatsal jain msc.pdf1.86 MBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.