Arithmetic Coding Context Grouping for Spectral Coefficients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for arithmetic coding face challenges in managing a large number of contexts and corresponding probability density functions, leading to increased encoding/decoding latency and memory requirements.
Innovation Solution
The method uses preceding spectral coefficients to determine context classes by non-uniformly quantizing their absolutes, reducing the number of contexts through grouping similar probability density functions into a single class, and employs a mapping to select an appropriate probability density function for encoding/decoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If a large neighbourhood of preceding coefficients is used for context-based arithmetic coding, then compression efficiency is improved, but the number of different possible contexts explodes combinatorially, leading to increased memory requirements and encoding/decoding latency
Solution Approach 1:
The patent merges multiple similar contexts into a single unified context by defining equivalence relations between context vectors. Contexts that differ only in insignificant ways are combined into the same context class, sharing a single probability density function. This dramatically reduces the total number of contexts from exponential growth to a manageable number, while preserving compression efficiency by maintaining distinctions only where they matter for prediction accuracy.
Solution Approach 2:
The patent transforms the context representation by applying parameter changes to the context vectors. Specifically, it uses hashing functions and quantization to map high-dimensional context vectors into a reduced parameter space. This parameter transformation allows the system to capture the essential characteristics of contexts without maintaining the full combinatorial complexity, thereby reducing memory requirements while preserving the predictive power needed for compression.
2Measurement precision
If more probability density functions are stored to handle different contexts, then coding precision is improved, but memory capacity requirements increase significantly
Solution Approach 1:
The patent combines multiple probability density functions into shared groups by merging contexts that are deemed equivalent. Instead of storing separate PDFs for every possible context combination, the system identifies contexts that can share the same PDF and merges them into unified context classes. This merging strategy maintains coding precision for distinct predictive patterns while dramatically reducing the total number of PDFs that must be stored in memory.
Solution Approach 2:
The patent implements universality by designing probability density functions that can serve multiple contexts simultaneously. A single PDF is designed to handle multiple equivalent contexts, making it multi-functional. This approach allows the system to achieve high coding precision across diverse contexts without proportionally increasing memory capacity, as one PDF performs the work of what would traditionally require many separate PDFs.
3Loss of information
If a huge number of probability density functions are managed, then compression performance is improved, but encoding/decoding latency increases
Solution Approach 1:
The patent merges contexts into equivalence classes to reduce the number of PDF lookups required during encoding and decoding. By combining similar contexts, the system reduces the search space from exponential to polynomial complexity. This merging strategy maintains compression performance by preserving the ability to distinguish meaningfully different contexts, while dramatically reducing the time required to select and apply the appropriate PDF during real-time encoding and decoding operations.
Solution Approach 2:
The patent performs preliminary action by pre-computing and storing only the essential context characteristics needed for PDF selection, rather than maintaining the full combinatorial space of contexts. During runtime, the system uses pre-established equivalence relations and hashing mechanisms to quickly identify the appropriate context class and associated PDF. This preliminary organization of context data enables fast retrieval and application during encoding/decoding, reducing latency while preserving compression performance.
Data Source
AI summary
The invention proposes a method and a device for arithmetic encoding of a current spectral coefficient using preceding spectral coefficients. Said preceding spectral coefficients are already encoded and both, said preceding and current spectral coefficients, are comprised in one or more quantized spectra resulting from quantizing time-frequency-transform of video, audio or speech signal sample values. Said method comprises processing the preceding spectral coefficients, using the processed preceding spectral coefficients for determining a context class being one of at least two different context classes, using the determined context class and a mapping from the at least two different context classes to at least two different probability density functions for determining the probability density function, and arithmetic encoding the current spectral coefficient based on the determined probability density function wherein processing the preceding spectral coefficients comprises non-uniformly quantizing absolutes of the preceding spectral coefficients for use in determining of the context class.