Multispectral simulation environment for modeling low-light-level sensor systems
Date
1998-11Author
Ientilucci, Emmett
Brown, Scott
Schott, John
Raqueno, Rolando
Metadata
Show full item recordAbstract
Image intensifying cameras have been found to be extremely useful in low-light-level (LLL) scenarios including military
night vision and civilian rescue operations. These sensors utilize the available visible region photons and an amplification
process to produce high contrast imagery. It has been demonstrated that processing techniques can further enhance the quality
ofthis imagery. For example, fusion with matching thermal JR imagery can improve image content when very little visible
region contrast is available. To aid in the improvement of current algorithms and the development of new ones, a high
fidelity simulation environment capable ofproducing radiometrically correct multi-band imagery for low-light-level
conditions is desired. This paper describes a modeling environment attempting to meet these criteria by addressing the task
as two individual components: (i) prediction ofa low-light-level radiance field from an arbitrary scene, and (ii) simulation of
the output from a low-light-level sensor for a given radiance field.
The radiance prediction engine utilized in this environment is the Digital Imaging and Remote Sensing Image Generation
(DIRSIG) model which is a first principles based multi-spectral synthetic image generation model capable of producing an
arbitrary number of bands in the 0.28 to 20 pm region. The DIRSIG model is utilized to produce high spatial and spectral
resolution radiance field images. These images are then processed by a user configurablemulti-stage low-light-level sensor
model that applies the appropriate noise and modulation transfer function (MTF) at each stage in the image processing chain.
This includes the ability to reproduce common intensifying sensor artifacts such as saturation and "blooming".
Additionally, co-registered imagery in other spectral bands may be simultaneously generated for testing fusion and
exploitation algorithms.
This paper discusses specific aspects ofthe DIRSIG radiance prediction for low-light-level conditions including the
incorporation ofnatural and man-made sources which emphasizes the importance ofaccurate BRDF. A description of the
implementation of each stage in the image processing and capture chain for the LLL model is also presented. Finally,
simulated images are presented and qualitatively compared to lab acquired imagery from a commercial system.