scispace - formally typeset
Search or ask a question
Topic

Light field

About: Light field is a research topic. Over the lifetime, 5357 publications have been published within this topic receiving 87424 citations.


Papers
More filters
Proceedings ArticleDOI
25 Jun 2001
TL;DR: This paper presents the wavelet stream which employs non-standard four-dimensional wavelet decomposition for Light Field compression, which allows for progressive transmission, storage, and rendering of compressed Light Field data.
Abstract: One of the most general image based object representations is the Light Field. Unfortunately, a large amount of data is required to reconstruct high quality views from a Light Field. In this paper, we present the wavelet stream which employs non-standard four-dimensional wavelet decomposition for Light Field compression. It allows for progressive transmission, storage, and rendering of compressed Light Field data. Our results show that 0.8% of the original coefficients or 0.3 bits per pixel, respectively are sufficient to obtain visually pleasing new views. Additionally, the wavelet stream allows for an adaptive multi-resolution representation of the Light Field data. Furthermore, a silhouetteencoding scheme helps to reduce the number of coefficients required. Our data structure allows to store arbitrary vector-valued data like RGB- or YUV-data. The Light Field data stored in the wavelet stream can be decompressed in real time for interactive rendering. For this, the reconstruction algorithm uses supplementary caching schemes.

24 citations

Proceedings ArticleDOI
04 Jul 2016
TL;DR: The proposed solution takes a noisy light field image and converts it to a sequence of epipolar images, using as intermediate step a representation based on the ordered sequence of the sub-aperture images, to reduce the effect of Gaussian noise.
Abstract: Many current light field cameras are based on a single sensor with an overlaid micro lenses array, making them more susceptible to noise. This paper proposes a novel light field denoising solution to reduce the effect of Gaussian noise as this is the most commonly assumed type of noise at the acquisition process. The proposed solution takes a noisy light field image and converts it to a sequence of epipolar images, using as intermediate step a representation based on the ordered sequence of the sub-aperture images. The created epipolar sequence is finally processed by a powerful, generic video denoising engine. The performance of the proposed denoising solution has been assessed using the PSNR and SSIM metrics for a representative set of rendered 2D views. The obtained results for two representative datasets compare favorably against state-of-the-art light field denoising methods, both in terms of objective assessment and visual appearance.

24 citations

Journal ArticleDOI
TL;DR: In this article, a phase mask is used to generate wavelets, which enable the creation of large-area deep sub-wavelength nanostructures by femtosecond laser irradiation onto various materials.
Abstract: The goal of creation of large-area deep sub-wavelength nanostructures by femtosecond laser irradiation onto various materials is being hindered by the limited coherence length. Here, we report solution of the problem by light field tailoring of the incident beam with a phase mask, which serves generation of wavelets. Direct interference between the wavelets, here the first-order diffracted beams, and interference between a wavelet and its induced waves such as surface plasmon polariton are responsible for creation of microgratings and superimposed nanogratings, respectively. The principle of wavelets interference enables extension of uniformly induced hybrid structures containing deep sub-wavelength nanofeatures to macro-dimension.

24 citations

Journal ArticleDOI
TL;DR: It is shown that learning an ensemble in the native dimensionality of the data promotes sparsity, hence increasing the compression ratio and sampling efficiency, and introducing a novel nonlocal pre-clustering approach that constructs an Aggregate MDE (AMDE).
Abstract: In this article we present a novel dictionary learning framework designed for compression and sampling of light fields and light field videos. Unlike previous methods, where a single dictionary with one-dimensional atoms is learned, we propose to train a Multidimensional Dictionary Ensemble (MDE). It is shown that learning an ensemble in the native dimensionality of the data promotes sparsity, hence increasing the compression ratio and sampling efficiency. To make maximum use of correlations within the light field data sets, we also introduce a novel nonlocal pre-clustering approach that constructs an Aggregate MDE (AMDE). The pre-clustering not only improves the image quality but also reduces the training time by an order of magnitude in most cases. The decoding algorithm supports efficient local reconstruction of the compressed data, which enables efficient real-time playback of high-resolution light field videos. Moreover, we discuss the application of AMDE for compressed sensing. A theoretical analysis is presented that indicates the required conditions for exact recovery of point-sampled light fields that are sparse under AMDE. The analysis provides guidelines for designing efficient compressive light field cameras. We use various synthetic and natural light field and light field video data sets to demonstrate the utility of our approach in comparison with the state-of-the-art learning-based dictionaries, as well as established analytical dictionaries.

24 citations

Journal ArticleDOI
TL;DR: This Letter considers passive Stokes parameter polarimeters and their inversion methods and finds that design of the null space of the inversion operator provides degrees of freedom to optimize the trade off between accuracy and signal-to-noise ratio.
Abstract: Imaging polarimeters infer the spatial distribution of the polarization state of the optical field as a function of time and/or wavelength. A polarimeter indirectly determines the polarization state by first modulating the intensity of the light field and then demodulating the measured data to infer the polarization parameters. This Letter considers passive Stokes parameter polarimeters and their inversion methods. The most widely used method is the data reduction matrix (DRM), which builds up a matrix equation that can be inverted to find the polarization state from a set of intensity measurements. An alternate strategy uses linear system formulations that allow band limited reconstruction through a filtering perspective. Here we compare these two strategies for overdetermined polarimeters and find that design of the null space of the inversion operator provides degrees of freedom to optimize the trade off between accuracy and signal-to-noise ratio. We further describe adaptive filtering techniques that could optimize the reconstruction for a particular experimental configuration. This Letter considers time-varying Stokes parameters, but the methods apply equally to polarimeters that are modulated in space or in wavelength.

24 citations


Network Information
Related Topics (5)
Optical fiber
167K papers, 1.8M citations
79% related
Image processing
229.9K papers, 3.5M citations
78% related
Pixel
136.5K papers, 1.5M citations
78% related
Laser
353.1K papers, 4.3M citations
78% related
Quantum information
22.7K papers, 911.3K citations
77% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
2023135
2022375
2021274
2020493
2019555
2018503