Modified SIFT Descriptors for Face Recognition under Different Emotions

Figure 1

(a) The middle point is known as candidate key point and orientations to this point are computed using pixel differences. (b) The value of each bin holds the magnitude sums from all the points computed within that orientation.

(a)

(b)