Inter-frame Predictive Coding Complexity Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing HEVC video coding standard has limited optimization methods for inter-frame predictive coding, leading to increased complexity and a need for additional optimization techniques to reduce coding complexity.
Innovation Solution
An inter-frame predictive coding method that determines maximum depths for coding tree units (CTUs) in reference and non-reference frames within a group of pictures, skipping certain mode decision processes and using specific algorithms like SAD and SATD to optimize the coding process, thereby reducing the complexity of inter-frame predictive coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If HEVC uses a hybrid coding framework with multiple innovative technologies (quadtree splitting, intra-frame prediction modes, advanced motion estimation, adaptive loop filter, semantic-based entropy coding), then compression efficiency is improved (50% bit rate reduction), but coding complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the CTU coding process into multiple depth levels (depth 0, 1, 2, 3) with different optimization strategies. At deeper depths (2 and 3), the patent segments out certain mode decision processes and uses simplified algorithms, while maintaining full processing at shallower depths where more complex decisions are still beneficial.
Solution Approach 2:
The patent changes the parameter of maximum CTU splitting depth from a fixed value to a dynamically determined value based on the reference frame type (P frame or B frame) and the closest already-coded reference frame. This parameter adjustment allows the system to adapt coding complexity to the specific coding situation, reducing complexity for B frames while maintaining efficiency for P frames.
2Manufacturing precision
If the CTU splitting depth is increased to improve coding precision, then compression performance is improved, but coding time increases
Solution Approach 1:
The patent applies partial action by performing complete mode decision processes only when necessary (at depth 0 and 1, and for P frames), while using simplified algorithms for partial processing at deeper depths (2 and 3) and for B frames. This selective application of full processing versus simplified processing reduces overall coding time while maintaining sufficient precision for the given frame type.
Solution Approach 2:
The patent implements skipping by omitting certain mode decision processes (specifically inter-frame N×2N and 2N×N modes) at depth 0 when the maximum depth is determined to be greater than 0. This skipping of redundant mode decisions at deeper splitting levels significantly reduces coding time without substantially impacting compression performance.
3Productivity
If all mode decision processes are performed for every CTU to ensure optimal coding, then coding performance is maximized, but processing complexity increases
Solution Approach 1:
The patent applies dynamics by making the processing complexity adaptive rather than static. The maximum CTU splitting depth and the set of modes to be decided are dynamically adjusted based on the frame type (P or B) and the position in the coding sequence. This dynamic adjustment allows the system to maintain high performance when needed while reducing complexity in situations where full processing is less critical.
Solution Approach 2:
The patent uses cheaper, simplified algorithms (such as direct mode selection without exhaustive mode decision) for B frames and at deeper CTU splitting depths, accepting that these are less optimal than full mode decision but sufficient for the given context. This trade-off reduces processing complexity significantly while maintaining acceptable coding performance for the specific frame types and depths involved.
Data Source
AI summary
The present invention discloses an inter-frame predictive coding method and a coder. Inter-frame predictive coding is sequentially performed on frames in multiple groups of pictures by using a group of pictures as a unit; a correlation between the reference B frame and a GPB frame in the same group of pictures is used; and when the maximum depth of the CTU in the same position in the GPB frame is smaller, a quantity of mode decisions performed on the CTU in the reference B frame may be relatively small, so that an objective of reducing complexity of an inter-frame predictive coding process of the reference B frame is achieved.