Combined Intra-Prediction for Lower-Residual Bitstream Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional intra-prediction methods in image encoding result in high residuals and higher bitrates due to the use of a single directional mode applied uniformly across a block, leading to inefficient encoding and decoding processes.
Innovation Solution
Implementing multi-directional intra-prediction within a block by applying different prediction modes and directions to different parts of the block, where prediction parameters such as modes and directions are inferred or interpolated from adjacent blocks, reducing the need for explicit signaling in the bitstream.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single directional mode is applied uniformly across a block, then the decoding process is simple and fast, but the prediction accuracy is low resulting in high residuals and higher bitrates
Solution Approach 1:
The current block is divided into multiple sub-blocks, each of which can be predicted using different intra-prediction modes and directions. This segmentation allows the decoder to achieve higher prediction accuracy by adapting to local variations in the image content, while the complexity is managed through systematic processing of each sub-block.
Solution Approach 2:
Different parts of the block (sub-blocks) are assigned different prediction modes and directions based on their local characteristics. This local quality approach ensures that each region is predicted using the most appropriate mode for its specific content, improving overall prediction accuracy without requiring complex global analysis.
2Loss of information
If multiple different prediction modes are applied to different parts of the block, then prediction accuracy improves and residuals are reduced, but the number of bits required to signal the prediction parameters increases
Solution Approach 1:
The prediction parameters for sub-blocks are determined based on the mode and direction used for the current block, establishing a preliminary framework before actual prediction. This preliminary action allows the decoder to infer many parameters without explicit signaling, reducing the bit overhead while maintaining multiple prediction modes.
Solution Approach 2:
The prediction parameters are derived self-service from already available information (the current block's prediction mode and direction) rather than requiring separate signaling. This self-service mechanism reduces the number of bits needed to transmit prediction parameters while still enabling multiple different modes across sub-blocks.
3Reliability
If prediction parameters are explicitly signaled in the bitstream, then the decoder can accurately reconstruct the predictions, but the bitrate increases
Solution Approach 1:
The decoder uses feedback from the current block's prediction mode and direction to determine the parameters for sub-block predictions. This feedback mechanism allows accurate reconstruction of predictions by deriving parameters from already-decoded information, eliminating the need for additional explicit signaling and thus reducing bitrate.
Data Source
AI summary
The present invention relates to a method of decoding in a decoder 1 a bitstream representing an image comprising: receiving the bitstream; for each block in the bitstream: forming predictions; and adding a respective prediction to each residual sample to form a reconstructed image sample; providing a reconstructed image from the reconstructed image samples; and outputting the reconstructed image; wherein the step of forming a prediction comprises: applying a first directional or non-directional prediction mode to reference samples to form at least a first set of one or more predictions for the current block; and applying one or more subsequent, different, directional or non-directional prediction modes to the same reference samples to form at least one subsequent set of one or more predictions for the current block, where any subsequent directional prediction mode differs from any first directional prediction mode in the spatial direction of prediction.