Video Encoding Control via Adaptive Re-encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding technologies face challenges in maintaining stable video quality while preventing CPB underflow, with 1-pass encoding leading to excessive bit rate suppression and quality degradation, and 2-pass encoding requiring significant computational resources.
Innovation Solution
The approach involves encoding video in units of encoding-order picture groups, where the encoding parameter is adjusted only when CPB underflow occurs, and re-encoding specific groups with increased quantization parameter step sizes or filter strength to reduce bit rate, while managing retry counts to minimize quality degradation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If 1-pass encoding is used to encode pictures sequentially, then encoding speed is improved, but CPB underflow suppression is insufficient and video quality degrades excessively
Solution Approach 1:
The patent performs preliminary buffer verification before actual encoding to predict CPB underflow conditions. By calculating the buffer status in advance using the hypothetical decoder model, the system can determine appropriate encoding parameters beforehand, allowing sequential encoding while preventing underflow without quality degradation.
Solution Approach 2:
The patent implements feedback by continuously monitoring the CPB residual bit rate during encoding and adjusting encoding parameters dynamically. When underflow is detected or predicted, the system adjusts quantization parameters or other encoding settings to reduce the bit rate, ensuring the CPB does not collapse while maintaining video quality.
2Reliability
If 2-pass encoding is used to encode all pictures twice, then CPB underflow suppression is improved, but computational resources and calculation amount increase significantly
Solution Approach 1:
The patent applies local quality by performing detailed buffer verification and adaptive re-encoding only for specific picture groups or frames where CPB underflow is detected or predicted, rather than encoding all pictures twice. This selective approach maintains reliable underflow suppression while significantly reducing overall computational resources compared to full 2-pass encoding.
Solution Approach 2:
The patent uses partial action by performing buffer verification and potential re-encoding only when necessary (when underflow is detected), rather than systematically processing all pictures through a second encoding pass. This reduces computational overhead while maintaining sufficient CPB underflow suppression.
3Manufacturing precision
If the quantization parameter is reduced to increase bit rate, then video quality is improved, but CPB underflow risk increases
Solution Approach 1:
The patent applies dynamics by making the quantization parameter adaptive rather than fixed. The encoding parameter is dynamically adjusted based on the current CPB residual bit rate and predicted underflow conditions. When the buffer is sufficient, higher quality parameters are used; when underflow risk is detected, the parameter is adjusted to reduce bit rate, ensuring continuous CPB stability.
Solution Approach 2:
The patent changes encoding parameters (such as quantization parameter, pre-filter strength) based on buffer status and predicted underflow conditions. By modifying these parameters dynamically during encoding, the system can optimize video quality while preventing CPB underflow, resolving the contradiction between quality and underflow prevention.
Data Source
Figure 1
Figure 2A~2D
Figure 3
AI summary
In a video encoding control method of the present invention, each picture is encoded in units of encoding-order picture groups such as GOPs, and when CPB underflow has occurred, an encoding parameter such as a quantization parameter and filter strength of a pre-filter is changed in such a manner that a generated bit rate is reduced, and an encoding-order picture group that is being encoded is re-encoded from its first picture using the changed encoding parameter.