A method of determining motion vectors.

In a method for determining motion vectors for an image signal divided into frames or partial images, wherein each image is divided into blocks and a motion vector is determined for each block in at least two stages, a number motion vector is assumed as possible in each stage and for each of these motion vectors, a difference function being respectively used for each block of the image point, based on the values ​​of which the most favorable vector is selected and is retained, optionally, as one of the vectors supposed as possible to a next stage, is provided for a maximum insensitivity to process noise in the first stage, subtracting a predetermined value from the value determined as a function of the deviation vector which has the shortest length and then one proceeds to comparison with the values ​​determined for e the deviation based on motion vectors of other assumed as possible and that in a following stage, subtracts a predetermined value every time the value of the deviation based on the motion vector, which in a preceding stage, has was determined as a vector of the most favorable movement, and comparing the thus reduced value of the difference function of this vector with the values ​​of the deviation function of other vectors of the same stage to determine the vector most favorable.