Directional Planar Intra Prediction for High-Efficiency Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image compression technologies face challenges in efficiently encoding and decoding high-resolution and high-quality images, particularly in handling intra prediction modes, especially for directional planar modes, which affect encoding efficiency and prediction performance.
Innovation Solution
The method and apparatus extend the planar mode to a directional planar mode, adaptively signal intra prediction mode information, derive reference samples, and apply PDPC to improve encoding efficiency by considering the correlation of residuals, using specific transform kernels for inverse transforms based on intra prediction modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional intra prediction modes (non-directional planar, horizontal, vertical) are used, then the encoding process is simple, but the prediction performance and encoding efficiency are insufficient for high-resolution and high-quality images
Solution Approach 1:
The patent segments the planar mode into two distinct directional planar modes: horizontal planar mode and vertical planar mode. This segmentation allows the encoding system to select the most appropriate directional mode based on the local image characteristics, thereby improving prediction accuracy without requiring a complete redesign of the intra prediction framework
Solution Approach 2:
The patent introduces dynamic mode selection mechanisms including PDPC (Prediction Directional Planar Mode Correction) and adaptive transform kernel selection. These dynamic elements allow the encoding system to adaptively adjust prediction parameters based on local image characteristics such as gradient directions and residual correlations, improving prediction performance while maintaining manageable encoding complexity through context-based adaptive processes
2Productivity
If directional planar mode is introduced to improve prediction accuracy, then encoding efficiency improves, but the complexity of signaling intra prediction mode information increases
Solution Approach 1:
The patent applies partial action by selectively signaling directional planar mode information only when beneficial. The planar direction flag is signaled adaptively based on whether horizontal or vertical planar mode is selected, rather than always signaling both directions. This reduces the average signaling overhead while still providing the capability to use directional planar modes when they improve encoding efficiency
Solution Approach 2:
The patent changes the parameter representation by introducing a planar direction flag that indicates whether horizontal or vertical planar mode is used, rather than using separate mode indices for each directional planar mode. This parameter change simplifies the signaling structure by using a binary flag instead of multi-bit mode identifiers, reducing signaling complexity while maintaining the ability to distinguish between different directional planar modes
3Manufacturing precision
If PDPC is applied to all prediction blocks, then prediction accuracy improves, but the processing time and computational complexity increase
Solution Approach 1:
The patent applies PDPC selectively based on local image characteristics rather than uniformly to all prediction blocks. The decision to apply PDPC is made adaptively based on local gradient analysis and residual correlation measurements, ensuring that PDPC processing is concentrated on regions where it provides the most benefit while avoiding unnecessary processing in regions where simple planar mode suffices
Solution Approach 2:
The patent implements partial application of PDPC by using it only for directional planar mode predictions where local image characteristics indicate benefit. The adaptive control mechanism applies PDPC to approximately 30-50% of prediction blocks based on local conditions, achieving significant prediction accuracy improvement while limiting the increase in processing time to only the necessary portions of the image data
4Loss of energy
If transform kernels are optimized for directional planar mode, then energy compaction improves, but the complexity of selecting and applying appropriate kernels increases
Solution Approach 1:
The patent changes the transform kernel selection parameter from a multi-level decision process to a simplified binary selection based on prediction mode. When directional planar mode is detected, a specific transform kernel (e.g., DST-II) is automatically selected, whereas traditional planar or directional modes use DCT kernels. This parameter change reduces kernel selection complexity by establishing a direct mapping between prediction mode type and transform kernel, eliminating the need for complex adaptive kernel selection algorithms
Solution Approach 2:
The patent establishes homogeneity in the transform kernel application by using consistent kernel types for specific prediction mode categories. All directional planar mode predictions use the same optimized transform kernel, and all traditional planar/directional modes use the same DCT kernel. This homogeneous approach simplifies the overall system by reducing the variety of kernel types and selection logic, making the system easier to implement while still achieving improved energy compaction for directional planar modes
Data Source
AI summary
An image decoding/encoding method and apparatus according to the present disclosure can derive an intra prediction mode of the current block from among pre-defined intra prediction modes, and generate a prediction block of the current block on the basis of the intra prediction mode. Here, the pre-defined intra prediction modes can include a non-directional planar mode, a directional planar mode, a horizontal mode and a vertical mode, and the directional planar mode can include a horizontal planar mode and/or a vertical planar mode.