Intra Prediction Using Temporal and Unreconstructed Reference Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding and decoding methods struggle to effectively utilize spatial and temporal reference samples for intra prediction, leading to inefficiencies in removing spatial and temporal redundancies in images.
Innovation Solution
The method involves determining and utilizing first and second sample sets, comprising reconstructed and unreconstructed samples adjacent to the current block, for intra prediction during image decoding and encoding, allowing for improved prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only reconstructed samples are used for intra prediction, then decoding order is simple and straightforward, but prediction accuracy is limited due to unavailable spatial and temporal reference samples
Solution Approach 1:
The patent applies preliminary action by performing reconstruction of certain blocks before using them as reference samples for intra prediction, even when those blocks would normally be decoded later in the standard decoding order. Specifically, the decoder reconstructs a second block before using it as a reference sample for predicting a first block, thereby making additional spatial and temporal reference samples available to improve prediction accuracy without fundamentally changing the overall decoding architecture
Solution Approach 2:
The patent extends the reference sample selection from the traditional spatial dimension only to include both spatial and temporal dimensions. By utilizing samples from previously decoded pictures (temporal references) in addition to neighboring spatial samples, the invention creates a multi-dimensional reference sample pool that significantly improves intra prediction accuracy, especially for inter-picture and intra-picture predictions
2Measurement precision
If unreconstructed samples are utilized for intra prediction, then prediction accuracy improves, but decoding order and process become more complex
Solution Approach 1:
The patent performs preliminary reconstruction of specific blocks (second blocks) before they would normally be decoded in the standard sequence, so that these blocks are ready to serve as reference samples when needed for predicting other blocks (first blocks). This preliminary action enables the use of unreconstructed samples without causing excessive delays in the overall decoding process
Solution Approach 2:
The patent applies local quality by selectively reconstructing only those specific blocks that will serve as reference samples for intra prediction, rather than reconstructing all blocks out of order. This localized approach to premature reconstruction minimizes the impact on overall decoding time while still providing the accuracy benefits of using unreconstructed samples where needed
3Adaptability or versatility
If traditional inter prediction with motion vectors is used, then temporal redundancy is removed, but intra prediction cannot leverage temporal reference samples effectively
Solution Approach 1:
The patent makes the intra prediction mechanism universal by enabling it to utilize both spatial reference samples (traditional intra prediction) and temporal reference samples (previously decoded blocks from other pictures). This multi-functionality allows the intra prediction to adapt to different prediction scenarios, improving its versatility and effectiveness across various video content types and prediction modes
Solution Approach 2:
The patent extends intra prediction from purely spatial reference sample utilization to include temporal dimension by incorporating samples from previously decoded pictures. This dimensional expansion allows intra prediction to leverage both spatial and temporal correlations in video data, significantly improving prediction accuracy while maintaining the simplicity of intra prediction processing
Data Source
AI summary
An image decoding method including determining at least one of a first sample set or a second sample set from samples reconstructed before a current block in a current picture or a previous picture. The first sample set corresponds to a first neighboring sample set adjacent to the current block. The second sample set corresponds to a second neighboring sample set adjacent to the current block. The first neighboring sample set comprises the samples reconstructed before the current block. The second neighboring sample set comprises samples that have not been reconstructed. The method includes performing intra prediction on the current block using a sample included in at least one of the first sample set or the second sample set as a reference sample of the current block. The method includes reconstructing the current block from a prediction block.