Skip to Main Content
Terahertz (THz) imaging is an innovative imaging technology that can provide a large amount of temporal and spectral information unavailable through other sensors. However, the huge amount and the relevance problem of features can be a barrier to analyze this type of images. In this study, we combine autoregressive and principal component analysis modeling to extract relevant features from the vast THz data sets. Afterward, K-harmonic-means clustering technique was used on the extracted features to segment THz images. Our approach of segmentation is evaluated by measuring performances and appraised by a comparison with some related works.
Date of Conference: 11-15 Nov. 2012