Fast Geometric Mode Decision in Video Encoders
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video encoding technologies, such as the MPEG-4 AVC Standard, face inefficiencies in capturing the geometric structure of 2D data, leading to sub-optimal video coding efficiency due to high computational complexity in geometric partitioning and reliance on simple statistics for mode decision.
Innovation Solution
The implementation of fast geometric mode decision algorithms in video encoders using geometric model-based partitioning modes, which include predictive search, hierarchical search, and alternate optimization of geometric partition and prediction data, to reduce computational complexity and improve coding efficiency by considering correlations between standard and geometric prediction modes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If geometric model-based partitioning modes are used to capture 2D data structure, then video coding efficiency is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by performing predictive search using neighboring block information before final mode decision. The encoder predicts geometric partition parameters (wedge angle, wedge distance) based on already decoded neighboring blocks, creating an initial candidate set that guides subsequent exhaustive search, thereby reducing the overall search space and computational complexity while maintaining coding efficiency
Solution Approach 2:
The patent segments the mode decision process into multiple stages: predictive search phase using neighboring information, candidate selection phase, and final exhaustive search phase. This segmentation allows the encoder to handle the complex geometric partitioning problem in manageable steps, reducing peak computational complexity while preserving the ability to find optimal modes
2Productivity
If exhaustive geometric mode testing is performed, then coding efficiency is maximized, but encoding time increases
Solution Approach 1:
The patent performs preliminary mode testing by evaluating geometric partition modes using predictive information from neighboring blocks before the final Rate-Distortion Optimization decision. This preliminary evaluation filters out clearly suboptimal modes early, allowing the encoder to skip exhaustive testing of unlikely candidates while maintaining thoroughness for promising modes, thus reducing encoding time without sacrificing coding efficiency
Solution Approach 2:
The patent applies partial action by performing exhaustive geometric mode testing only for blocks where predictive search indicates potential benefit. For blocks where neighboring information suggests standard partitioning is sufficient, the encoder performs limited or no geometric mode testing, thereby reducing average encoding time across all blocks while maintaining optimal performance where needed
3Measurement precision
If geometric partitioning is used to capture object boundaries, then representation accuracy is improved, but device complexity increases
Solution Approach 1:
The patent introduces an intermediary predictive search mechanism that uses neighboring block information to estimate geometric partition parameters before final encoding decisions. This intermediary step acts as a bridge between simple standard partitioning and complex exhaustive geometric mode testing, providing accurate representation where needed while reducing complexity through intelligent prediction based on spatial correlations in video data
Data Source
Figure 1
Figure 2~4
Figure 3A~3B
AI summary
There are provided methods and apparatus for fast geometric mode decision in a video encoder. An apparatus includes an encoder (100) for performing a non-exhaustive geometric mode decision using geometric model-based partitioning modes. The encoder (100) skips a testing of the geometric model-based partitioning modes based on at least one of a correlation and a dependence between the geometric model-based partitioning modes and non-geometric partitioning modes.