Page 133 - ITU Journal, ICT Discoveries, Volume 3, No. 1, June 2020 Special issue: The future of video and immersive media
P. 133
ITU Journal: ICT Discoveries, Vol. 3(1), June 2020
entropy encoded transform coefficients. residuals. These residuals form the starting point
The amount depends on the chosen type of for the encoding process of the first enhancement
transform (nLayers). Additionally, if the temporal sub-layer. A number of coding tools, which will be
prediction is used, for each processed plane an described further in the following subsection,
additional chunk with temporal data for process the input and generate entropy encoded
Enhancement sub-layer 2 is present. quantized transform coefficients.
5. CODING STRUCTURE 5.1.3 Enhancement sub-layer 2
As a last step of the encoding process, the
5.1 Encoder
enhancement data for Layer 2 needs to be
The encoding process to create an LCEVC generated. In order to create the residuals, the
conformant bitstream is shown in Fig. 2 and can be coefficients from Layer 1 are processed by an in-
depicted in three major steps. loop decoder to achieve the corresponding
reconstructed picture. Since Layer 1 might have a
5.1.1 Base codec different resolution than the input sequence, the
Firstly, the input sequence is fed into two reconstructed picture is processed by an upscaler,
consecutive downscalers and is processed again depending on the chosen scaling mode.
according to the chosen scaling modes. Any Finally, the residuals are calculated by a subtraction
combination of the three available options of the input sequence and the upscaled
(2-dimensional scaling, 1-dimensional scaling in the reconstruction.
horizontal direction only or no scaling) can be used. Similar to Layer 1, the samples are processed by a
The output then invokes the base codec which few coding tools. Additionally, a temporal
produces a base bitstream according to its own prediction can be applied on the transform
specification. This encoded base is included as part coefficients in order to achieve a better removal of
of the LCEVC bitstream. redundant information. The entropy encoded
5.1.2 Enhancement sub-layer 1 quantized transform coefficients of Layer 2, as well
as a temporal layer specifying the use of the
The reconstructed base picture may be upscaled to temporal prediction on a block basis, are included
undo the downscaling process and is then in the LCEVC bitstream.
subtracted from the first-order downscaled input
sequence in order to generate the Layer 1 (L-1)
Encoder Configuration Headers
Entropy Temporal
Encoding Layer
L-2 residuals
Input - Temporal Transform Quantization Entropy L-2 Coefficient
Sequence Prediction Encoding Layers
Upscaler
Downscaler Inverse Inverse
+ L-1 Filter Transform Quantization
L-1 residuals
- Transform Quantization Encoding L-1 Coefficient
Entropy
Layers
Downscaler Upscaler
reconstruction
Base
Encoder bitstream Encoded Base
Fig. 2 – Structure of an LCEVC encoder
© International Telecommunication Union, 2020 111