METHOD FOR FORMING PREDICTION PICTURES

摘要

In conventional moving-image encoders, image areas are displaced block by block and it is not taken into account whether moving or unmoving image content exists within such blocks. In the process according to the invention, moving image contents are detected with the aid of an object mask and displaced independently of the surrounding unmoving image background. Instead of a block-oriented predicted image, an object-oriented predicted image is determined. For this purpose, shape and type of the moving image content is determined in the receiver from the motion information normally transmitted and the conventionally decoded image sequence and stored in an object mask which continuously tracks moving objects. Using this object mask, an improved display image can be generated by copying the image background from an image store. The method according to the invention leads to a qualitatively improved reproduction of coded moving-image sequences. In addition, intermediate images can be generated in which moving image content is also displaced independently of the image background.