Scalable video coding supporting pixel value refinement scalability

A more efficient way of enabling scalability in terms of pixel value resolution is achieved by temporally predicting (126, 142) a first and a second representation of a video material separately to each other with the first representation being of a lower pixel value resolution than the second representation, with mapping the first prediction residual or a reconstructed version thereof from the first pixel value resolution to the second pixel value resolution dependent on the second prediction signal obtained from temporally predicting the second representation, and by coding (132, 134, 144) a second prediction residual being of the second pixel value resolution as representing a deviation between a combination of the second and third prediction signals and the second representation.