Parallel CABAC Decoder Using MPS Path Estimation for Higher Throughput
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AVC/H.264 CABAC decoders face limitations in throughput and parallelism, which restrict their efficiency in high-profile applications and resolutions, such as 1920×1088 at 64 frames per second, and do not effectively support higher standards without compromising coding performance.
Innovation Solution
A high throughput parallel AVC/H.264 context-based adaptive binary arithmetic coding decoder is designed using a top-down methodology and dataflow models, leveraging the similarity with Viterbi decoders to increase parallelism by estimating and decoding most probable symbol paths, thereby enhancing output rate without decreasing coding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the degree of parallelism of the CABAC decoder is increased, then the output rate and throughput are improved, but the device complexity and difficulty of implementation increase
Solution Approach 1:
The patent applies preliminary action by estimating the most probable symbol (MPS) path in advance before actual decoding. The decoder predicts which symbols are most likely to occur and pre-calculates their decoding paths, allowing parallel processing of multiple probable paths simultaneously. This reduces the need for sequential processing while managing complexity through intelligent prediction rather than brute-force parallelism.
Solution Approach 2:
The patent segments the decoding process into multiple independent parallel paths based on different symbol probability estimates. Each processing element handles a specific segment of the bitstream corresponding to a particular probable symbol path, allowing simultaneous processing of multiple segments. This segmentation enables throughput improvement while keeping individual processing elements relatively simple.
2Speed
If the degree of parallelism of the CABAC decoder is increased, then the output rate is improved, but the difficulty of detecting and measuring increases
Solution Approach 1:
The patent implements feedback mechanisms where the decoder continuously monitors the actual symbol values decoded and compares them with the predicted MPS paths. Based on this feedback, the system dynamically adjusts which parallel paths are followed and how resources are allocated. This feedback loop simplifies control by allowing the system to adapt to actual conditions rather than requiring complex pre-planning for all possible scenarios.
Solution Approach 2:
The decoder performs preliminary estimation of MPS paths to identify which parallel processing paths are most likely to be needed. This preliminary analysis allows the system to prepare and organize parallel processing resources in advance, making the actual parallel execution more manageable and easier to control by focusing computational effort on the most probable paths.
3Productivity
If parallel decoding paths are estimated and decoded beforehand, then the throughput is improved, but the loss of information or coding performance may deteriorate
Solution Approach 1:
The patent applies partial action by estimating and preparing only the most probable symbol paths rather than all possible paths. The decoder focuses computational resources on the top few most likely paths based on probability estimates, performing partial parallel decoding instead of exhaustive parallel decoding. This maintains coding performance by concentrating effort on the most relevant paths while achieving throughput improvement through parallel processing of these selected paths.
Solution Approach 2:
The patent changes parameters dynamically by adjusting which paths are pursued based on probability estimates and actual decoded values. The system modifies its processing focus in real-time, shifting resources between different parallel paths as needed. This parameter adaptation ensures that coding performance is maintained by following the most probable paths while still benefiting from parallel processing throughput improvements.
Data Source
AI summary
Methods for decoding are provided. The proposed method includes steps of: receiving a most probable symbol (MPS) value and a probability value for generating a probability model update; and receiving the probability model update for generating the MPS value and the probability value, wherein when the probability value shows that an MPS is occurred, a path corresponds to the MPS is estimated and a first bin included in the path is decoded beforehand.