Possibility-Based Video Coding for Non-Stationary Sources
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional video coding methods based on stationary probability distribution are not applicable to non-stationary video sources, leading to inefficiencies in coding due to unstable frequency distributions and inaccurate probability estimation.
Innovation Solution
An information preserving coding method that estimates the possibility distribution of a sample's value using prediction operators and exponential smoothing, generating a bitstream based on this distribution to improve coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional probability distribution-based coding methods are used, then coding can be performed for stationary sources, but coding efficiency deteriorates for non-stationary video sources with unstable frequency distributions
Solution Approach 1:
The patent applies dynamics by transitioning from static probability distribution models to dynamic possibility distribution models that adapt to changing video content characteristics. The possibility distribution is updated continuously based on current context, allowing the coding system to respond to non-stationary patterns in video sources while maintaining coding efficiency.
2Measurement precision
If probability estimation is performed for non-stationary sources, then coding can be attempted, but prediction accuracy deteriorates due to unstable frequency distributions
Solution Approach 1:
The patent changes the fundamental parameter from probability (which assumes stationarity) to possibility distribution (which accommodates non-stationarity). This parameter transformation allows the system to maintain high prediction accuracy for video sources with unstable frequency distributions by not relying on the unstable probability estimates.
3Productivity
If Shannon's coding theory is applied, then optimal code rate is achieved for stationary sources with known distribution, but the method becomes inapplicable when probability distribution is unknown or non-stationary
Solution Approach 1:
The patent implements self-service by enabling the coding system to automatically adapt to unknown and changing distributions without requiring prior knowledge or manual intervention. The possibility distribution is derived directly from the video content itself through contextual analysis, allowing the system to serve itself in determining appropriate coding parameters for any given source.
Data Source
AI summary
The invention provides an information preserving coding method based on possibility distribution of value of a sample. The possibility distribution of value of a sample refers to evaluation of the possibility of various values of the sample. The information preserving coding method provided is more in line with the non-stationary probability property of an actual source, and a coding code length is the overhead caused by a prediction error of the possibility distribution of value of the sample. The invention further provides multiple coding methods based on possibility distribution of value of a sample. Through improving the prediction accuracy of the possibility distribution of value of the sample, the coding efficiency is greatly improved.