Image

M. Gouiffès, V. Belissen, H. Bull, H. Chaaban, E. Frenoux, F. Vernier

The image is seen here both as a source of information to be interpreted (image in input) and as a medium for visualizing information (image in output).

  • Extracting visual information from images and videos
  • Image and video recognition and analysis
  • Information visualization
  • Computer vision for camera-projector systems

Extracting visual information from images and videos

image1

  • Estimation of the optical flow, definition of image and video primitives and descriptors (trajectory flow), visual salience based on color and movement, evaluation of the relevance of the visual attributes.
  • Detection of geometric structures from moving depth-color sensors and / or from the optical flow.

Methods : points of interest, estimation of optical flow, fuzzy logic, data fusion, region descriptors by covariance, level lines, clustering, methods by vote (UV-velocity)

Image and video recognition and analysis

image2

Recognition of :

  • movement, human activities, signs and linguistic structures in the Sign Language
  • videos places (for locating robots)
  • crystals or combinations of crystals (for medical applications)

Methods : machine learning, deep learning (LSTM recurrent networks), clustering

Information Visualization

image3

Information visualization is a mature and established field for many years. The Image topic contributes to :

  • invent new visualizations
  • apply visualization techniques to realistic datasets
  • to confront interaction techniques with the real needs of the user in the workplace

This work is applied to geolinguistic data (regional languages), finance, weather, economics, transport, sport.
Methods : GapCharts, Stratum Graphs, Braided Graphs, Heatmaps, Voronoi contours, Horizon maps, scatterplots.

Computer vision for camera-projector systems

image4 

  • Projected augmented reality: how to project images onto colored or textured surfaces (even dynamic and non-planar) while perceiving the same images as those projected on an ideal screen (white and plane) ?
  • Structured light approaches: characterization and reconstruction in real time of surfaces

Methods : Methods: geometric and photometric calibration, profilometry by Fourier transform, phase sequence.

 

LIMSI
Campus universitaire bât 507
Rue du Belvédère
F - 91405 Orsay cedex
Tél +33 (0) 1 69 15 80 15
Email

SCIENTIFIC REPORTS

Scientific report

LIMSI in numbers

8 Research Teams
100 Researchers
40 Technicians and Engineers
60 Doctoral Students
70 Trainees

 

Paris-Saclay University new window


Logo DataIA

 

© 2017 LIMSI CNRS