15 November 2007 PCA-based visualization of terahertz time-domain spectroscopy image
Author Affiliations +
Proceedings Volume 6787, MIPPR 2007: Multispectral Image Processing; 67871M (2007) https://doi.org/10.1117/12.749995
Event: International Symposium on Multispectral Image Processing and Pattern Recognition, 2007, Wuhan, China
A novel visualization method of terahertz time-domain spectroscopy (THz-TDS) image is presented, which is based on principal component analysis (PCA) technique. The proposed method include three processing steps: firstly, the THz- TDS image is preprocessed using a spatial vector filtering technique to denoise. Secondly, the THz-TDS image is transformed from spatio-temporal domain to spatio-spectral domain, and the transformed image can be viewed as a multispectral image whose spectral dimensionality D is equal to the sampled number of THz-TDS pulse at each pixel. Thirdly, each of spectrum vector at a pixel is viewed as a point in D dimensional space, the covariance matrix of pixels can be computed, and then three eigenvectors corresponding to the first 3 largest eigenvalues are found by PCA technique. the THz-TDS image is projected along these three eigenvectors. By normalizing these 3 principal component images and mapping them into the RGB space, we can get a synthetic color image as a visualization result of the THz- TDS image. Due to vector-based dimensionality reduction, the proposed method can provide more visual information of the THz-TDS image than scalar-based visualization techniques. Finally, experimental results are provided to demonstrate the performance of the proposed method.
© (2007) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jihong Pei, Jihong Pei, Yong Hu, Yong Hu, Weixin Xie, Weixin Xie, "PCA-based visualization of terahertz time-domain spectroscopy image", Proc. SPIE 6787, MIPPR 2007: Multispectral Image Processing, 67871M (15 November 2007); doi: 10.1117/12.749995; https://doi.org/10.1117/12.749995

Back to Top