Intra Prediction Modes for Video Compression Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video codecs encode images using limited intra prediction modes, which reduces their efficiency in compressing high-resolution or high-quality video content, leading to increased calculation and bit rate requirements.
Innovation Solution
The method involves using intra prediction modes with various directivities, allowing for more flexible prediction directions by employing (dx, dy) parameters to improve image compression efficiency and reduce calculation during the prediction process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If limited intra prediction modes are used in conventional video codecs, then device complexity is reduced, but image compression efficiency deteriorates
Solution Approach 1:
The patent implements dynamic selection of intra prediction modes based on block characteristics. The system adaptively chooses between planar, DC, and directional prediction modes depending on the specific block being encoded, rather than using a fixed limited set or always using the most complex mode. This dynamic adaptation improves compression efficiency while managing device complexity through conditional logic.
Solution Approach 2:
The patent changes the parameter of prediction mode selection by introducing multiple directional modes (varying the direction parameter) in addition to planar and DC modes. This parameter expansion allows the system to better match different image characteristics, improving compression efficiency. The system manages complexity by providing coded syntax elements that efficiently signal the selected mode parameter.
2Productivity
If more intra prediction modes with various directivities are used, then image compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments the intra prediction functionality into distinct mode categories: planar mode, DC mode, and directional modes. Each segment handles specific types of image content effectively. This segmentation allows the system to achieve high compression efficiency by selecting the appropriate segment for each block while keeping overall device complexity manageable through modular organization of prediction functions.
Solution Approach 2:
The patent creates a universal intra prediction framework that can handle various image characteristics through multiple modes. The same prediction framework universally applies to all blocks, selecting from planar, DC, or directional modes based on content analysis. This multi-functionality improves compression efficiency across diverse content types while avoiding the need for separate specialized systems for each mode type.
3Power
If conventional intra prediction modes are used, then calculation amount is reduced, but prediction accuracy deteriorates
Solution Approach 1:
The patent performs preliminary analysis of block characteristics before selecting the prediction mode. By examining gradient information and variance in advance, the system determines the most suitable prediction mode for each block. This preliminary action enables accurate prediction without requiring exhaustive calculation of all possible modes, thus maintaining low computational overhead while improving prediction accuracy.
Solution Approach 2:
The patent incorporates feedback mechanisms where the encoding process evaluates the effectiveness of different prediction modes and uses this information to guide mode selection. The system analyzes the residual error and block characteristics, providing feedback that informs the choice of prediction mode for subsequent blocks or refinement of current predictions. This feedback loop improves prediction accuracy while managing calculation through intelligent decision-making.
Data Source
Figure 1
Figure 2~4
Figure 5
AI summary
A method and apparatus for encoding and decoding an image through intra prediction using a pixel of the neighboring block along an extended line having a predetermined gradient about a pixel inside the current block.