Biologically inspired enhancement of dim light video
Malm, Henrik LU ; Oskarsson, Magnus LU
and Warrant, Eric
LU
(2012)
p.71-85
- Abstract
- In this chapter a technology for the enhancement of video data obtained at low light levels is presented. The method was inspired by the way in which nocturnal animals adaptively sum intensities, spatially and temporally, to improve vision at night. Due to the low photon count under these conditions the visual input is dark and unreliable, which leads to noisy low contrast images. The noise becomes very apparent when we try to enhance the contrast and, by this, amplify the intensities in the darkest regions of the images. By constructing spatio-temporal smoothing kernels that automatically adapt to the three dimensional intensity structure at every point, the noise can be considerably reduced, with fine spatial detail being preserved and... (More)
- In this chapter a technology for the enhancement of video data obtained at low light levels is presented. The method was inspired by the way in which nocturnal animals adaptively sum intensities, spatially and temporally, to improve vision at night. Due to the low photon count under these conditions the visual input is dark and unreliable, which leads to noisy low contrast images. The noise becomes very apparent when we try to enhance the contrast and, by this, amplify the intensities in the darkest regions of the images. By constructing spatio-temporal smoothing kernels that automatically adapt to the three dimensional intensity structure at every point, the noise can be considerably reduced, with fine spatial detail being preserved and enhanced without added motion blur. For color image data, the chromaticity is restored and demosaicing of raw RGB input data can be performed simultaneously with the noise reduction. The method is a very generally applicable one, contains only few user-defined parameters and has been developed for efficient parallel computation using a graphics processing unit (GPU). The technique has been applied to image sequences with various degrees of darkness and noise levels. Results from some of these tests, and comparisons to related work, are presented here. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/5052476
- author
- Malm, Henrik
LU
; Oskarsson, Magnus
LU
and Warrant, Eric
LU
- organization
- publishing date
- 2012
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Frontiers in Sensing: From Biology to Engineering
- editor
- Barth, Friedrich G. ; Humphrey, Joseph A. C. and Srinivasan, Mandyam V.
- pages
- 71 - 85
- publisher
- Springer
- external identifiers
-
- scopus:84930018492
- ISBN
- 978-3-211-99749-9
- 978-3-211-99748-2 (Print)
- DOI
- 10.1007/978-3-211-99749-9_5
- language
- English
- LU publication?
- yes
- id
- 310769b2-b76a-4084-9c24-eb5463870b25 (old id 5052476)
- date added to LUP
- 2016-04-04 11:09:53
- date last changed
- 2022-01-29 21:25:39
@inbook{310769b2-b76a-4084-9c24-eb5463870b25,
abstract = {{In this chapter a technology for the enhancement of video data obtained at low light levels is presented. The method was inspired by the way in which nocturnal animals adaptively sum intensities, spatially and temporally, to improve vision at night. Due to the low photon count under these conditions the visual input is dark and unreliable, which leads to noisy low contrast images. The noise becomes very apparent when we try to enhance the contrast and, by this, amplify the intensities in the darkest regions of the images. By constructing spatio-temporal smoothing kernels that automatically adapt to the three dimensional intensity structure at every point, the noise can be considerably reduced, with fine spatial detail being preserved and enhanced without added motion blur. For color image data, the chromaticity is restored and demosaicing of raw RGB input data can be performed simultaneously with the noise reduction. The method is a very generally applicable one, contains only few user-defined parameters and has been developed for efficient parallel computation using a graphics processing unit (GPU). The technique has been applied to image sequences with various degrees of darkness and noise levels. Results from some of these tests, and comparisons to related work, are presented here.}},
author = {{Malm, Henrik and Oskarsson, Magnus and Warrant, Eric}},
booktitle = {{Frontiers in Sensing: From Biology to Engineering}},
editor = {{Barth, Friedrich G. and Humphrey, Joseph A. C. and Srinivasan, Mandyam V.}},
isbn = {{978-3-211-99749-9}},
language = {{eng}},
pages = {{71--85}},
publisher = {{Springer}},
title = {{Biologically inspired enhancement of dim light video}},
url = {{http://dx.doi.org/10.1007/978-3-211-99749-9_5}},
doi = {{10.1007/978-3-211-99749-9_5}},
year = {{2012}},
}