The method includes the step of forming differences between a current image section (Sf (k, l) ) and a preceding image section (Sf-1 (k+i, l+j) ) for a number of pixel vectors (k, 1) of the image sections and for a number of movement vectors (i, j). The absolute value of a respective difference is associated with a quantised difference (T (k, l, i, j) ), corresponding to a stepped quantisation characteristic with exponentially rising level width and exponentially rising level height. All quantised difference values or all squared quantised difference values are added to a sum value (LPDC (i,j) ) to form a motion vector for each of the number of pixel vectors. The most probable motion vector is determined, by selecting the motion vector with the lowest sum value.
展开▼