A luminance signal (Y) is applied to a segmentation device (2) and subjects the signal to a segmentation algorithm to produce a region image signal which is encoded in a region image encoder (3). The region image signal is also applied to a region post-processor (4) in which an adapted smooth function is added to the region signal. The resulting modified region image signal is subtracted in subtractor (5) from the luminance signal to produce a texture signal which is encoded by a texture image encoder (6). The encoded signals are conveyed via a data channel (9) to respective region (12) and texture (14) image decoders and the recovered region image signal post-processed in a processor (13) in the same manner as with processor (4). The resulting modified region image signal and the recovered texture image signal are added in an adder (15) to produce a reconstructed luminance signal. <IMAGE>