Karr, Brian (2019) Accurate reproduction of real-world lighting. PhD thesis, University of Warwick.

Preview

PDF WRAP_Theses_Karr_2019.pdf - Submitted Version - Requires a PDF viewer. Download (102Mb) | Preview

Request Changes to record.

Abstract

The dynamic range of a scene is defined as the difference in lighting between the brightest and darkest areas of that scene. The goal of High Dynamic Range (HDR) imaging is to capture all of the detail in a scene, no matter what the dynamic range. HDR is essential when capturing safety critical environments, for example during a rocket launch. Here it is crucial to be able to see simultaneously all the detail in the rocket plume, on the rocket itself and within the launch environment. Although the capture of HDR images and video has been well studied, little is known about the physical accuracy of the captured image compared to the scene itself. Just how truly representative is the captured HDR image of the real world lighting? In this thesis, the image capture pipeline is analysed in detail, while not assuming linearity, in order to maintain physical accuracy throughout the workflow. Rather than approximating the camera response function, calibrated measurements are employed to characterize both the image sensor and the associated optics.

As a first step in this research, existing HDR capture technology was investigated. A number of latest generation digital imaging systems were selected for testing and evaluation, and several limitations are identified in the evaluation of existing HDR capture technology. First, in most cases, no standard is utilized when stating imager dynamic range capabilities. Second, in-camera processing of data is described generically, such as with a vendor specific log curve applied, but only in some cases is the processing described, or the reverse transform provided in order to linearize the data. Third, there is lack of discussion, or available data, related to acceptable noise level or the noise level used in the determination of signal to noise based sensor characterizations. From the imager test and evaluation, it was concluded that the processing pipelines differ significantly from manufacturer to manufacturer, and that there is no clear definition as to how imaging system dynamic range is either measured or specified.

To address these limitations, a new method for characterizing HDR response from a \system" perspective was developed, including the combined effect of the optics, the sensor hardware, and the camera processing. The results indicated that the accuracy of manufacturer stated capabilities are dependent upon the chosen maximum RMS noise defining the minimum luminance step used to determine dynamic range. Different manufacturers appear to use different metrics, making comparison difficult.

The model was further improved to treat the optical and sensor elements separately, so that the contributions from each are known. To provide maximum accuracy, the model in this thesis categorizes error via methodical calibration; including determination of sensor noise attributes, system gain response, and the effects of lens glare. Additionally, using reference measurements from a calibrated luminance meter, absolute radiometric calibration was achieved, albeit with a degree of error. The results highlighted that the effects of glare limit the ability to accurately capture and measure the lower luminance portion of a scene. To capture an HDR image most accurately, glare effects must be reduced by using higher quality optics, or by reducing sources of glare in the scene.

Item Type:	Thesis (PhD)
Subjects:	T Technology > TR Photography
Library of Congress Subject Headings (LCSH):	High dynamic range imaging, Lighting -- Computer simulation, Glare
Official Date:	2019
Dates:	Date Event 2019 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	Warwick Manufacturing Group
Thesis Type:	PhD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	Chalmers, Alan ; Debattista, Kurt, 1975-
Format of File:	pdf
Extent:	xi, 136 leaves : illustrations
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item

Downloads