Image transform for video coding

A method for decomposing two disjoint sets of pictures, including a first set and a second set formed from a given set of input pictures, into a set of low-band pictures and a set of high-band pictures respectively in an image decomposing unit having at least one prediction motion compensator and at least one update motion compensator, said prediction motion compensator and update motion compensator being motion compensators that shift blocks or segments of a picture according to an associated set of motion vectors,

• wherein a picture in the second set is predicted by at least two prediction motion compensators which use a same picture in the first set to generate at least two different prediction motion-compensated signals, and said picture in the first set is chosen individually for said picture in the second set, the method comprising:
• calculating a high-band picture by subtracting from the same corresponding picture in the second set the at least two different prediction motion-compensated signals which are generated from said same picture in the first set, the prediction motion-compensated signals being scaled for subtraction;
• and
• forming a low-band picture by adding to the same corresponding picture in the first set at least two different update motion-compensated signals which are generated from a same high-band picture, the update motion-compensated signals being scaled for addition,
• wherein the at least two different update motion-compensated signals are generated from the same high-band picture that received the at least two different prediction motion-compensated signals from said same picture in the first set.