Inter-component prediction

A method for decoding a multi-component picture spatially sampling a scene with respect to different components, comprising:

• reconstructing a first component signal relating to a first component of the multi-component picture from a data stream;
• and
• reconstructing a portion of a second component signal relating to a second component of the multi-component picture based on a spatially corresponding portion of the reconstructed first component signal, a correction signal derived from the data stream, or both,
• wherein the first and second components are color components and the first component signal is a prediction residual associated with the first component and the second component signal is a prediction residual associated with the second component, the decoding of the multi-component picture comprising block-based hybrid video decoding including:
• sub-dividing the multi-component picture into tree blocks, subdivide the tree blocks using recursive multi-tree subdivisioning into code blocks, subdivide each code block into prediction blocks and into residual blocks, and subdivide the residual blocks into transform blocks;
• selecting prediction modes based on first information from the data stream;
• determining prediction parameters for the prediction blocks based on second information from the data stream;
• deriving a prediction signal using the prediction modes and the prediction parameters;
• deriving a residual signal within each residual block by performing inverse transformations within the transform blocks;
• reconstructing the multi-component picture by combining the prediction signal and the residual signal;
• and
• extracting, from the data stream, a signaling syntax element that indicates, at a granularity of the transform blocks, switching between (a) reconstructing of the second component signal from the spatially corresponding portion of the reconstructed first component signal and the correction signal, and (b) reconstructing the second component signal from the correction signal independent of the spatially corresponding portion of the reconstructed first component signal.

An encoder configured to encode a multi-component picture spatially sampling a scene with respect to different components, by

• encoding a portion of a second component signal relating to a second component of the multi-component picture based on a spatially corresponding portion of a reconstructed first component signal, a correction signal, or both;
• and
• inserting the correction signal into the data stream,
• wherein the first and second components are color components and the first component signal is a prediction residual associated with the first component and the second component signal is a prediction residual associated with the second component, the encoder comprising a block-based hybrid video encoder configured to:
• sub-divide the multi-component picture into tree blocks, subdivide the tree blocks using recursive multi-tree subdivisioning into code blocks, subdivide each code block into prediction blocks and into residual blocks, and subdivide the residual blocks into transform blocks;
• encode, into the data stream, first information representing prediction modes;
• encode, into the data stream, second information representing prediction parameters for the prediction blocks;
• derive a prediction signal using the prediction modes and the prediction parameters;
• determine a residual signal for each residual block based on the prediction signal;
• and
• encode, into the data stream, the residual signal by performing transformations on the residual signal,
• wherein the encoder is configured to encode, into the data stream, a signaling syntax element that indicates, at a granularity of the transform blocks, switching between (a) encoding the second component signal from the spatially corresponding portion of the reconstructed first component signal and the correction signal, and (b) encoding the second component signal from the correction signal independent of the spatially corresponding portion of the reconstructed first component signal.