摘要 |
Visual indices (IBi) of the image are converted (2) into observed indices (IOBi) in a chosen system of representation. These indices (IOBi) are later set into correspondence (5) with a predicted index (IPRi), thus enabling each observed index (IObi) to be assigned an identification label (INSi). This label is assigned to the corresponding input visual index (IBi), the evolution of which can be logged from image to image. Updating (3) of tracked indices (IMi) is carried out with the aid in particular of a Kalman prediction, and the new indices (IPRi) are predicted (4) for the next image. <IMAGE> |