MOTION VECTOR DETECTOR FOR DYNAMIC IMAGE, MOTION COMPENSATION PREDICTOR USING THE SAME AND INTER-MOTION COMPENSATION PREDICTION FRAME ENCODER

摘要

PURPOSE:To detect a motion vector at a lattice point itself with high accuracy for the low bit encoding of a dynamic image. CONSTITUTION:It is assumed that transmission transformations vx=a+bx+cy+dx<2>+exy and vy=f+gx+hy+ey<2>+dxy are used in compensation prediction. Such equations are accumulated in a motion model accumulation part 24 as the model information 24a of the motion vector, and the motion vector at each lattice point can be found from the constraint condition of the space-time gradient of a picture element value and the transformation equations by calculation in a motion vector calculation part 26 by using a space derivative image 21a and a time derivative image 22a.