主权项 |
1. An apparatus for performing inter-layer prediction between a first picture coding stage for coding a first tone-mapped version of a picture, comprising a first dynamic sample value range, and a second picture coding stage for coding a second version of the picture, comprising a second dynamic sample value range greater than the first dynamic sample value range, the apparatus comprising:
a global predictor configured to derive a global tone-mapping function based on a statistical analysis of pairs of values of co-located samples in the first tone-mapped version and the second version of the picture, and apply the global tone-mapping function onto the first tone-mapped version of the picture so as to achieve a globally predicted reference picture for the second version of the picture; and a local predictor configured to locally derive a locally varying tone-mapping function based on a statistical analysis of values of co-located samples in the second version of the picture and the globally predicted reference picture in units of sub-portions into which the globally predicted reference picture and the second version of the picture are partitioned, and apply the locally varying tone-mapping function onto the globally predicted reference picture so as to achieve a globally and locally predicted reference picture for the second version of the picture, wherein the local predictor is configured to, in locally deriving the locally varying tone-mapping function, for each sub-portion, determine a slope and an intercept of a regression line through a distribution of pairs of values of co-located samples in the respective sub-portion of the second version of the picture and the globally predicted reference picture,and encode the slope and the intercept determined for the sub-portions as side information into an enhancement datastream generated by the second picture coding stage, and forming an enhancement layer relative to a datastream output by the first picture coding stage, and
wherein the second picture coding stage is configured to lossy code a prediction residual between the globally and locally predicted reference picture and the second version of the picture into an enhancement datastream, and the local predictor is configured toencode the slope determined for the respective sub-portion as a prediction residual of a spatial prediction from the slope of a neighboring sub-portion the slope of which has been previously encoded,fit a straight line comprising the slope of the respective sub-portion, into a distribution of pairs of values of co-located samples in the neighboring sub-portion of the second version of the picture as far as reconstructible based on the prediction residual and the globally and locally predicted reference picture on the one hand, and the globally predicted reference picture on the other hand and determine the intercept of the straight line; andencode the intercept of the respective sub-portion as a prediction residual relative to a spatial intercept prediction depending on the intercept of the straight line. |