Abstract: In a method of video image motion compensation a plurality of motion vectors are derived by techniques such as phase correlation between pairs of images. In order to produce a set of motion vectors which can be used to define substantially all the motions in an input image, sets of vectors are derived from comparisons with preceding and succeeding images and are then combined to produce a set of vectors for assignment to an input image. Once the set of vectors has been derived they are assigned to the input image from where they can be projected to produce a desired output image. The vectors are assigned to individual elemental areas of the image and certain elemental areas have two (or more) vectors assigned to them.

Abstract: Non-linear picture transformations in a video image motion compensation system are caused by camera zooms or rotations. Sets of trial vectors are derived for each block of the image representing peak correlation values between adjacent images in the sequence. A constant representing the non-linear transformation is estimated using the sets of trial vectors and this enables a vector representing the non-linear transformation to be added to the trial vector list for each block.