OPtics and Topography Involving the Cornea And Lens

optical-list@cs.berkeley.edu

Vision-Realistic Rendering

Introduction

We introduces a new concept, that of vision-realistic rendering. New rendering techniques are developed for the computer generation of synthetic images that incorporate accurate optics, especially for the human visual system, both as a model and from specific patient data. Two approaches are developed. In the first approach, we convolve an image with a point spread function, using high dynamic range source images to achieve a much more realistic estimate of glare and loss of contrast. The second approach traces rays from a retina or film plane through a system of lenses and into a three-dimensional scene. Both of these approaches can be used for camera models, eye models, and actual corneal topographic data. We are incorporating an accurate distribution of photoreceptors, with the density of sampling decreasing away from the foveal region of the retina. There are two distinct impacts of this research, one from the perspective of computer graphics and the other from the point of view from optometry and ophthalmology.

Papers and Publications

Brian A. Barsky, Billy P. Chen, Alexander C. Berg, Maxence Moutet, Daniel D. Garcia, Stanley A. Klein, "Incorporating Camera Models, Ocular Models, and Actual Patient Eye Data for Photo-Realistic and Vision-Realistic Rendering," Abstract in the Fifth International Conference on Mathematical Methods for Curves and Surfaces, June 29 - July 4, 2000, Oslo, Norway. [pdf] [ps]
Brian A. Barsky, Daniel D. Garcia, Stanley A. Klein, Woojin M. Yu, Billy P. Chen, Sarang S. Dalal, "RAYS (Render As You See): Vision-Realistic Rendering Using Hartmann-Shack Wavefront Aberrations," internal report, March 2001. [pdf] [ps] [high res images]

Projects

Modelling the Human Eye using BMRT (Spring 2000)
In modelling the human eye , we investigate light rays that enter through the cornea, are refracted by the lens, and then contact the retina. We use the Renderman spec. to model the eye and we use BMRT to ray trace the light rays that are not obstructed by the iris.
Camera/Ocular Raytracing (Summer 2000)
Our improved approach traces rays from the retina or film plane through an optical system of lenses (ocular or camera-based) and computes radiances of rays permeating through a three-dimensional scene. The illumination from the scene along a ray can be determined either from a synthetic computer-modeled environment or from data captured by a series of images taken in a real environment. In the synthetic case, the light can be computed using traditional ray tracing or an experimental virtual light field technique. For a real scene, the flow of light is represented using a light field or Lumigraph representation of the plenoptic function.
CWhatUC : Software Tools for Predicting, Visualizing and Simulating Corneal Visual Acuity (Dr. Daniel Garcia's Ph.D. thesis, 2000)
The cornea is the transparent tissue covering the front of the eye, and performs about two-thirds of the refraction, or bending of light into the eye. Thus, subtle variations in its shape significantly affect a patient's visual acuity. Clinicians need to know the shape and refractive contribution of the cornea for several reasons: corneal refractive surgery, contact lens fitting and diagnosis of eye conditions. Instruments that measure the cornea's topography are called corneal topographers (CTs), and they have recently become a quite common tool for clinicians. These devices typically shine rings of light onto the cornea and capture the reflection pattern with a video camera. We extract the raw data from the CT and then we construct a spline surface representation from these reflection patterns. All our visualization and analysis is performed on these corneal surface representations. In this work, we present CWhatUC, a set of software tools for the prediction, visualization and simulation of corneal visual acuity.
RAYS (Render As You See): Vision-Realistic Rendering Using Hartmann-Shack Wavefront Aberrations
RAYS (Render As You See) is a system for "vision-realistic rendering" which can simulate the vision of actual individuals. Patient data is derived from sensors, called Shack-Hartmann devices, that capture the wavefront aberrations that are present in the patient's entire optical pathway in response to a single point source of light in the retina.
The input to RAYS is an image, corresponding depth information, and a wavefront derived from the Shack-Hartmann device. Given a focusing distance, our vision-realistic rendering algorithm then blurs the scene appropriately. The result is an image that closely approximates what the actual patient would have seen if focused at that distance.
Vision-realistic rendering is particularly interesting in the context of laser refractive eye surgeries such as PRK and LASIK. Currently, almost a million Americans per year are choosing to undergo such elective surgeries. RAYS could convey to doctors the vision of a patient before and after surgery. In addition, RAYS could provide accurate and revealing visualizations of predicted acuity and simulated vision to potential candidates for such surgeries to facilitate educated decisions about the procedure. Still another application would be to show such candidates the possible visual anomalies that could arise from the surgery (such as glare at night).

Presentations

"RAYS (Render As You See): Vision-Realistic Using Hartmann-Shack Wavefront Aberrations," Biggraph (Berkeley Siggraph group), UC Berkeley, February 2001.