Spectral imaging of human portraits and image quality

Show full item record

Title: Spectral imaging of human portraits and image quality
Author: Sun, Qun
Abstract: This dissertation addresses the problem of capturing spectral images for human portraits and evaluating image quality of spectral images. A new spectral imaging approach is proposed in this dissertation for spectral images of human portraits. Thorough statistical analysis is performed for spectral reflectances from various races and different face parts. A spectral imaging system has been designed and calibrated for human portraits. The calibrated imaging system has the ability to represent not only the facial skin but also the spectra of lips, eyes and hair from various races as well. The generated spectral images can be applied to color-imaging system design and analysis. To evaluate the image quality of spectral imaging systems, a visual psychophysical image quality experiment has been performed in this dissertation. The spectral images were simulated based on real spectral imaging system. Meaningful image quality results have been obtained for spectral images generated from different spectral imaging systems. To bridge the gap between the physical measures and subjective visual perceptions of image quality, four image distortion factors were defined. Image quality metrics were obtained and evaluated based statistical analysis and multiple analysis. The image quality metrics have high correlation with subjective assessment for image quality. The image quality contribution of the distortion factors were evaluated. As an extension of the work other researchers in MCSL have initiated, this dissertation research will, working with other researchers in MCSL, put effort to build a publicly accessible database of spectral images, Lippmann2000.
Record URI: http://hdl.handle.net/1850/15387
Date: 2003-07

Files in this item

Files Size Format View
QSunThesis07-2003.pdf 27.73Mb PDF View/Open

The following license files are associated with this item:

This item appears in the following Collection(s)

Show full item record

Search RIT DML


Advanced Search

Browse