Video Intra Prediction Using DIMD and ISP for Lower Signaling Overhead
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in reducing signaling overhead and improving coding efficiency, particularly in intra prediction modes, which are essential for efficient video processing.
Innovation Solution
The proposed solution combines decoder-side intra prediction mode derivation (DIMD) with intra prediction using subpartitions (ISP) to derive at least one intra prediction mode (IPM), thereby reducing signaling overhead and enhancing coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional intra prediction modes are used, then coding complexity is maintained, but signaling overhead increases and coding efficiency decreases
Solution Approach 1:
The decoder derives intra prediction modes autonomously using DIMD algorithms without requiring explicit mode signaling from the encoder. The decoder uses local gradient information and template matching to self-determine the most probable intra prediction modes, eliminating the need for mode index transmission and reducing signaling overhead.
Solution Approach 2:
The patent extracts only the essential gradient information and template data from the video block at the decoder side, deriving intra prediction modes from this extracted information rather than transmitting complete mode information. This extraction approach reduces the amount of data that needs to be signaled while maintaining prediction accuracy.
2Manufacturing precision
If more intra prediction modes are signaled, then prediction accuracy improves, but signaling overhead and bitrate increase
Solution Approach 1:
The patent derives multiple intra prediction modes (including planar, DC, and angular modes) partially and selectively at the decoder side based on gradient analysis, rather than signaling all possible modes. This partial derivation approach achieves sufficient prediction accuracy for most blocks without the bitrate cost of signaling every mode.
Solution Approach 2:
The patent changes the parameter representation from explicit mode indices to implicit gradient-based mode derivation. By transforming the problem from selecting from a discrete set of modes to continuously deriving modes from gradient information, the system achieves higher prediction accuracy without increasing bitrate, as the gradient information is already available from the decoded block.
3Productivity
If intra prediction with subpartitions is used, then coding efficiency improves, but signaling overhead increases
Solution Approach 1:
The patent divides video blocks into subpartitions (e.g., 2x2, 4x2, 2x4 partitions) and applies DIMD independently to each subpartition. This segmentation allows the decoder to derive tailored intra prediction modes for each subregion, improving coding efficiency by capturing local variations while avoiding the need to signal modes for the entire block, thus reducing overall signaling overhead.
4Loss of information
If decoder-side intra prediction mode derivation is applied, then signaling overhead reduces, but computational complexity at decoder increases
Solution Approach 1:
The patent performs gradient calculation and template preparation as preliminary actions during the decoding process, before the actual mode derivation. By pre-computing these intermediate values from the decoded block, the subsequent mode derivation steps can proceed more efficiently, reducing the peak computational power required at the decoder while maintaining the signaling overhead benefits of DIMD.
Data Source
AI summary
Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. The method comprises: processing, during a conversion between a picture of a video and a bitstream of the video, a first video block in the picture based on a combination of decoder-side intra prediction mode derivation (DIMD) and intra prediction with subpartitions (ISP), the DIMD being used for deriving at least one intra prediction mode (IPM); and performing the conversion based on the first video block. Compared with the conventional solution, the proposed method can advantageously improve the coding effectiveness and coding efficiency.