An Improved colorimetric invariants and RGB-Depth-based Codebook model for background subtraction using kinect
Résumé
In this paper we propose to join the benefits of multiple in- variant information into the well-know background subtraction method 'Codebook'. Indeed, this method mainly repose on a color model allowing a separate process of color and intensity distortion. In order to manage hard situations involving high illumination changes, we propose to enhance this model with the use of two supplementary steps: 1/ transforming the input color image using a colorimetric invariant in order to obtain a color-invariant image whatever the illumination conditions; 2/ using depth information as a new data inside the Codebook model, thus performing an RGB-D fusion during the segmentation process.
Origine | Fichiers produits par l'(les) auteur(s) |
---|