METHOD AND DEVICE FOR PROCESSING A SEQUENCE OF IMAGES IN ORDER TO DETERMINE CONTINUOUS THUMBNAILS IN SAID SEQUENCE OF IMAGES

The present disclosure relates to a method for processing a sequence of images acquired by a vision sensor of a scene, the method comprising at least one step of identifying, in an image referred to as the reference image, object positions, independently of the content of the reference image, and a step of tracking the objects on the basis of the object positions in the reference image in each of a plurality of images, referred to as the tracking images, acquired after the reference image, the step of tracking the objects in each processed tracking image comprising: determining a predicted overall registration transformation between the reference image and the processed tracking image; predicting object positions in the processed tracking image according to the object positions in the reference image and the predicted overall registration transformation for the processed tracking image; searching for the objects in the processed tracking image on the basis of the predicted object positions.