Sung Ho Park

Abstract

In recent years, the graphics community is seeing an increasing demand for the capture and usage of high-dynamic-range (HDR) images. Since the production of HDR imagery is not solely the domain of the visualization of real life or computer generated scenes, novel techniques are also required for imagery captured from non-visual sources such as remote sensing, medical imaging, astronomical imaging, etc. This research proposes to integrate the techniques used for the display of high-dynamic-range pictorial imagery for the practical visualization of non-pictorial (scientific) imagery for data mining and interpretation. Nine algorithms were utilized to overcome the problem associated with rendering the high-dynamic-range image data to low-dynamic-range display devices, and the results were evaluated using a psychophysical experiment. Two paired-comparison experiments and a target detection experiment were performed. Paired-comparison results indicate that the Zone System performs the best on average and the Local Color Correction method performs the worst. The results show that the performance of different encoding schemes depend on the type of data being visualized. The correlation between the preference and scientific usefulness judgments (R2 = 0.31) demonstrates that observers tend to use different criteria when judging the scientific usefulness versus image preference. The experiment was conducted using observers with expertise (Radiologists) for the Medical image to further elucidate the success of HDR rendering on these data. The results indicated that both Radiologists and Non-radiologists tend to use similar criteria regardless of their experience and expertise when judging the usefulness of rendered images. A target detection experiment was conducted to measure the detectability of an embedded noise target in the Medical image to demonstrate the effect of the tone mapping operators on target detection. The result of the target detection experiment illustrated that the detectability of targets the image is greatly influenced by the rendering algorithm due to the inherent differences in tone mapping among the algorithms.

Library of Congress Subject Headings

Imaging systems--Data processing--Evaluation; Color vision--Computer simulation

Publication Date

8-1-2004

Document Type

Thesis

Department, Program, or Center

School of Photographic Arts and Sciences (CIAS)

Advisor

Montag, Ethan

Advisor/Committee Member

Fairchild, Mark

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK8315 .P37 2004

Campus

RIT – Main Campus

Download

Share

COinS