Wavefront Coding Methods for Extended Depth of Field Image Acquisition and Display
Akpinar, Ugur (2023)
Tampere University
2023
Tieto- ja sähkötekniikan tohtoriohjelma - Doctoral Programme in Computing and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Väitöspäivä
2023-09-29
Julkaisun pysyvä osoite on
https://urn.fi/URN:ISBN:978-952-03-3051-4
https://urn.fi/URN:ISBN:978-952-03-3051-4
Tiivistelmä
Depth of field is a fundamental factor that influences the performance of imaging systems. Limited depth of field in image sensing significantly reduces the quality of spatial information outside the range of focus—an effect referred to as defocus blurring. Limited focusing characteristics of contemporary near-eye three-dimensional displays create inaccurate depth perception and subsequently cause visual discomfort due to conflicting visual cues. Most notable is the rivalry between vergence and accommodation, referred to as the vergence-accommodation conflict.
This dissertation adopts the principle of computational extended depth of field as a means to solve the aforementioned issues in image sensing and in near-eye displays by establishing the link between the vergence-accommodation conflict and the display depth of field. Specifically, the dissertation presents two novel solutions based on wavefront coding and implemented through a combination of refractive lens and diffractive phase mask to engineer the corresponding system point spread functions and make them depth-invariant.
A key principle in the presented methodology is the co-design of the optical elements and the image processing algorithms. To this end, we have developed physically accurate differentiable image formation models of computational cameras and near-eye displays and combined them with modern machine-learning techniques to learn novel optical elements together with the parameters of the neural processing modules to jointly optimise the targeted extended depth-of-field imaging systems. We have made a theoretical contribution about how to decide the search space for the phase masks to be optimised and have integrated the viewer’s optics and important characteristics of the human visual system into the new display design.
An extensive set of simulations analysed both quantitatively and qualitatively, has demonstrated that our method outperforms state-of-the-art techniques. We have further validated the developed methods through experiments on benchtop prototypes of the designed systems.
This dissertation adopts the principle of computational extended depth of field as a means to solve the aforementioned issues in image sensing and in near-eye displays by establishing the link between the vergence-accommodation conflict and the display depth of field. Specifically, the dissertation presents two novel solutions based on wavefront coding and implemented through a combination of refractive lens and diffractive phase mask to engineer the corresponding system point spread functions and make them depth-invariant.
A key principle in the presented methodology is the co-design of the optical elements and the image processing algorithms. To this end, we have developed physically accurate differentiable image formation models of computational cameras and near-eye displays and combined them with modern machine-learning techniques to learn novel optical elements together with the parameters of the neural processing modules to jointly optimise the targeted extended depth-of-field imaging systems. We have made a theoretical contribution about how to decide the search space for the phase masks to be optimised and have integrated the viewer’s optics and important characteristics of the human visual system into the new display design.
An extensive set of simulations analysed both quantitatively and qualitatively, has demonstrated that our method outperforms state-of-the-art techniques. We have further validated the developed methods through experiments on benchtop prototypes of the designed systems.
Kokoelmat
- Väitöskirjat [4848]