I s a a c  C O H E N 



 M i s c e l l a n e o u s 




Invention Disclosures

  • "Detection and Tracking of moving objects in a video stream" . (more details ) (Office of Technology Licensing file #2940).
  • "Motion stabilization and Mosaicing"
    The motion compensation and mosaicing algorithm we have developed address the problem of estimation of the camera motion from the video stream acquired by the camera without any additional knowledge. We present a feature-based approach with robust motion estimation by controlling the distribution of the feature points. The camera motion model is represented by an affine transformation. The parameters of this affine transform are derived by searching for the set of affine parameters that has the largest support amongst the extracted set of matching points.
    (Office of Technology Licensing file #3358).
  • "Detection of Moving Objects from a moving platform by defining a dynamic background model"
    The background model is built by considering a set of registered frames and characterizing for each image pixel a distribution model representing the pixel color changes within the sliding window. We use several distribution models to characterize if a pixel is moving.
    (Office of Technology Licensing file #3359).
  • "Inference of Human Postures by training and classification of 3D body reconstruction"
    An appearance-based, view-independent, 3D shape description for classifying and identifying human posture using Machine Learning techniques. The proposed global shape description is invariant to rotation, scale and translation and varies continuously with 3D shape variations. (U.S. Patent Pending)
    (Office of Technology Licensing file #3378).
  • "3D Human Body Tracking from multiple synchronized video sensors using an articulated body model"
    This technique uses a mixture of Gaussians approach which allows unknown static (shape) parameters to be estimated simultaneously with dynamic (pose) parameters within the particle filtering scheme. Auxiliary measurements are used to improve the sampling distribution and to estimate subsets of state parameters deterministically. These auxiliary measurements include optical flow cues and the detection of individual body parts. The proposed approach eliminates the need for manual initialization of the body pose and body shape before tracking. (U.S. Patent Pending)
    (Office of Technology Licensing file #3379).





  • "CGI/Perl et JavaScript, La création de pages HTML interactives" Editions Eyrolles. 1996 (in french)
  • "CGI/Perl y JavaScript, La creatione de pages HTML interactives" Editions Eyrolles. 1997 (in spanish)