Method and apparatus for the avoidance of repeated random errors in transformation coefficients of image signals moved

摘要

A method of quantizing coefficient blocks of DCT coefficients representing a motion picture signal in a manner that prevents repetitive mismatch errors. The coefficient blocks include a current coefficient block derived from a current picture block having a position in a current picture, and a previous coefficient block derived from a previous picture block located in a previous picture at the position of the current picture block in the current picture. The current picture is a P-picture, the previous picture is either the P-picture or the I-picture processed immediately preceding the current picture. In the method, the previous coefficient block is quantized using a first step size. A second step size for quantizing the current coefficient block is determined. The second step size is set to be different from the first step size. Finally, the current coefficient block is quantized using the second step size set in the setting step.