| 摘要 |
The invention relates to a method for annotating, with a set of K landmarks, each image I i of a set of images to annotate, from a set of previously annotated images. According to the invention, said method is implemented by a processor and comprises, for each annotated face image I j :
- determining (11) a warping between said image I i and said image I j ,
- warping (12) said image I i using said warping delivering a warped image
- determining (13) image similarity information Sim(i,j) between said image I j and said warped image
and:
- selecting (14), among said set of previously annotated face images, a best annotated image I jbest being associated with similarity information of the highest value,
- determining (15) a location (x,y) of each landmark L i,k of said set of K landmarks of said image I i by using a corresponding location of each landmark of said best annotated image I jbest . |
