Enhanced Night Vision Via a Combination of Poisson Interpolation and Machine Learning

Abstract

Our research initiative focuses on enhancing semiconductor-based night-vision imagery via The fundamental problems of imaging under low-light conditions are noise suppression, overcoming low image contrast, and the loss of important perceptual cues. We have developed three new image-processing techniques to address these problems. These include non-linear spatio-temporal denoising filters, variants of Poisson integration for localizedcontrast adjustment, and machine-learning methods for reintroducing perceptual cues. The goal of our work is to enhance night-vision imagery by improving its sensitivity, and aiding in its rapid and accurate interpretation. We have applied our methods to low-light visible, near infrared (NIR), and short-wave infrared images (SWIR). In this annual report on the first phase of our research, we describe and evaluate the processing methods that we have developed to date. Our objective in this phase was to demonstrate the feasibility and potential of these new computational approaches.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Feb 01, 2006
Accession Number
ADA451258

Entities

People

  • Leonard Mcmillan
  • Wei Wang

Organizations

  • University of North Carolina at Chapel Hill

Tags

Communities of Interest

  • Advanced Electronics
  • Autonomy
  • C4I
  • Sensors

DTIC Thesaurus Topics

  • Air Force
  • Artificial Intelligence
  • Cameras
  • Computational Science
  • Computer Graphics
  • Computer Science
  • Computer Vision
  • Detection
  • Digital Images
  • Image Processing
  • Infrared Detectors
  • Intellectual Property
  • Machine Learning
  • Pattern Recognition
  • Short-Wavelength Infrared Radiation
  • Three Dimensional
  • Visible Spectra

Readers

  • Aviation Science / Aeronautics.
  • Economics
  • Image Processing and Computer Vision.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks
  • Microelectronics