Government ApplicationsCollaborative research effort with NAVAIR-WD, China Lake, CA and ONR
 

Cognitech, Inc. has developed unique, state-of-the-art technology for universal, multi-channel Video-Registration (e.g. Visual vs. IR), Frame-Fusion Super-Resolution, mosaic reconstruction, real-time video-stabilization, database video-geo-registration and tracking. Unlike any of the existing registration methods, Cognitech’s algorithms are contrast-invariant, meaningful level-set geometry-based, and threshold-less.

UNIVERSAL MULTI-CHANNEL VIDEO-REGISTRATION  (VIDEO TO IR)

gov_infrared_image.jpg 2.jpg 3.jpg 4.jpg
Infrared image Visible image Unregistered visible image superimposed on IR image (mismatched geometry is observed)  
IR to Visible cross-registration

 

Cognitech’s patented video-registration technology is unlike any other registration methods used presently in industrial and military applications. This technology does not rely on image brightness, as in correlation-based methods, nor does it rely on ad-hoc edge-detection, and pyramidal multi-scale analysis that is not geometrically-invariant (as in wavelet-based analysis).

Cognitech’s break-through image and video registration technology explores intrinsic geometrical structures embedded into the image data. This geometrical information is invariant of color and brightness changes common in image acquisition, due to changes in weather, time-of-day, season, camouflage, clutter, etc.

Since the intrinsic image geometries are invariant under various imaging modes, Cognitech’s registration works even across sensor modalities, e.g. visual, IR, LADAR, etc.

GOVERNMENT REGISTRATION
PROJECTIVE-INVARIANT, AUTOMATIC, MOSAIC RECONSTRUCTION AND SUPER-RESOLUTION

1.jpg 2.jpg 3.jpg
Video Frame Consecutive Video Frame Geometrically-Accurate Mosaic

 

VIDEO-GEO-REGISTRATION
 5.jpg

1. Projective registration of a video-stream on-to Geo-Tiff Data-Base

2. Matching a mobile object (a blue container) appearing in the video and not in the reference to the Geo-Tiff image

3. Computing accurate geo-location coordinates of the movable object