Keynotes
Light Fields at Disney Research
Speaker: Alexander Sorkine-Hornung
Abstract
Sensors capture the world at ever higher spatial, temporal and angular resolution. For instance, modern cinema cameras record 120 fps at 4k resolution and beyond. Consequently, one obvious problem is data management and complexity. However, for computer vision and graphics research, a more fundamental challenge arises. In the past we have carefully tuned our models and algorithms to address issues of sparse sampling, from basic image processing tasks to image-based 3D reconstruction and optical flow. In this talk I will argue that, for very densely sampled input such as light fields, these existing strategies do not necessarily get the best out of the data and can even be counter-productive. Extremely dense sampling poses interesting novel research challenges and opportunities beyond simply improving speed and scalability of existing techniques, since various concepts, ranging from local edge detection to global optimization, become meaningless or fundamentally impractical to compute. I will present our ongoing work at Disney Research on 3D scene and object modeling from light fields, which breaks with a number of established practices in image-based reconstruction in order to achieve reconstruction quality and speed unparalleled by existing methods.
Pelican's Light Field Camera Arrays: Performance and Applications
Speaker: Kartik Venkataraman
Abstract
As camera arrays transition from the research labs to commercial/mobile imaging devices it is interesting to understand the limits of their performance both in terms of the resolutions that they can scale to and the quality of the depth maps that they can support. There are many factors that determine the final resolution of the camera array imaging system's optical format, F#, number of cameras in the array, pixel crosstalk and pixel MTF, and the superresolution factor. We revisit the superresolution factor from the point of view signal aliasing and set limits in terms of minimum and maximum observable frequencies to understand its impact on the final achievable resolution. We also review the accuracy of the depth maps achievable from the camera array and highlight the key system parameters that define it. Finally, we illustrate practical applications from the use of the camera array in day to day use and how it might facilitate a different user experience than the one people are currently used to.
Light Field Cameras for Industrial Applications
Speaker: Christian Perwass
Abstract
Light Field cameras are a new type of 3D-cameras that capture a standard image together with the depth information of a scene with a single camera through a single lens in a single shot using just the available light. Raytrix has specialized on developing light field cameras for industrial applications. A patented micro lens array design allows for an optimal compromise between high effective resolution and large depth of field. Raytrix cameras are already in use in applications like volumetric velocimetry, plant phenotyping, automated optical inspection and microscopy, to name a few.
The talk will introduce the technology, give an overview of the algorithms for refocusing, depth estimation and metric calibration, and show many industrial applications.
Glass-Free 3D Light Field Displays
Speaker: Gordon Wetzstein
Abstract
In this talk, we explore modern approaches to glasses-free 3D display using compressive light field displays. In contrast to conventional technology, compressive displays aim for a joint-design of optics, electronics, and computational processing that together exploit compressibility of the presented data. For instance, multiview images or light fields show the same 3D scene from different perspectives - all these images are very similar and therefore compressible. By combining displays that use multilayer architectures or directional backlighting combined with optimal light field factorizations, limitations of existing devices, for instance resolution, depth of field, and field of view, can be overcome. In addition to light field display, we will discuss approaches to compressive super-resolution image display and compressive high dynamic range display. As with compressive light field displays, these technologies rely on multiplexing image content in time such that the visual system of a human observer combines presented patterns into a consistent 3D, high-resolution, or high-contrast image. With the invention of integral imaging and parallax barriers in the beginning of the 20th century, glasses-free 3D displays have become feasible. With rapid advances in optical fabrication, digital processing power, and computational perception, a new generation of display technology is emerging: compressive displays exploring the co-design of optical elements and computational processing while taking particular characteristics of the human visual system into account. We will review these techniques and also give an outlook on next-generation compressive light field camera technology.