The Color Image team designs automatic analysis schemes for images that represent the color information in observed scenes. Its researchers work on
- Extracting local or global features from CFA or raw images acquired by color cameras equipped with one sensor,
- Selecting color spaces that are adapted to the segmentation of full color images that have been provided by three-sensor cameras,
- Recognizing objects under uncontrolled acquisition conditions by the analysis of multi-spectral images in the visible spectrum.
Here is our main application: The quality or dimension control of manufactured objects. For example:
color vision system which detects aspect flaws occurring on the color surfaces of drinking glasses decorated thanks to an industrial silk-screen process
embedded vision system which measures the cross section of catenary contact wire
color vision system that inspects the frozen meat.
We also develop automatic diagnostic of turbo-alternator by pattern recognition with L2EP Institute.