Video coding using temporally coherent dynamic range mapping

摘要

A more efficient co-use of dynamic range mapping on the one hand and temporal prediction on the other hand such as, for example, in order to code HDR frame sequences, is achieved by exploiting the concept of weighted prediction in order to transition from the mapping parameter from the reference frame to the currently temporally predicted frame. By this measure, the temporal prediction does not fail and despite the frame-wise variation in the dynamic range mapping, encoding efficiency is, thus, maintained. As a favorable side aspect, weighted temporal prediction is already within the capabilities of existing video coding stages such as, for example, the H.264/AVC.

主权项

1. Video decoder for reconstructing a first sequence of frames, the sample values of which are represented in a first format covering a first dynamic range, from a data stream, comprising
a video decoding stage configured to reconstruct, from the data stream, a second sequence of frames the sample values of which are represented in a second format covering a second dynamic range lower than the first dynamic range, by a weighted temporal prediction of a first frame of the second sequence using a second frame of the second sequence, weighted by a weighting parameter and offset by an offset parameter comprised by the data stream, as a reference; a parameter setter configured to set a mapping parameter for the first frame of the second sequence depending on a mapping parameter for the second frame of the second sequence, the weighting parameter and the offset parameter, and a sample value reconverter configured to convert the sample values of the frames of the second sequence from the second format to the first format using a mapping function which maps the second dynamic range onto a portion out of the first dynamic range which is set by the mapping parameter for the respective frame of the second sequence.