Video Encoding Mode Switching Without Pipeline Retracing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video image encoding devices with pipeline structures face challenges in managing code generation within a specific maximum value, leading to decreased processing speed and complex retracing of processes when the IPCM mode is frequently switched, especially in HEVC and H.264 encoding.
Innovation Solution
A video image encoding device that includes a predictive encoder, subtracter, predictive residual encoder and decoder, adder, header code string generator, and coefficient code string generator, which allows for variable-length-coding of residual coefficients or directly using residual decoded images as code strings, depending on modes, to control code generation without retracing the pipeline.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If the pipeline is retraced when switching to IPCM mode to suppress the amount of code, then the code amount can be controlled, but the processing speed decreases and the encoding process may not complete in required time
Solution Approach 1:
The system performs preliminary determination of whether to use residual coding or PCM mode before the pipeline reaches the variable-length coding stage. By deciding the coding mode in advance (at the orthogonal transformation/quantization stage), the system avoids the need to retrace the pipeline when code amount exceeds the limit, thereby maintaining processing speed while controlling code quantity.
Solution Approach 2:
The system dynamically switches between residual coding mode and PCM mode based on real-time code amount feedback. When the code amount approaches the maximum limit, the system transitions to PCM mode which directly uses pixel values without complex processing, allowing flexible adaptation to different image contents while maintaining the encoding deadline.
2Quantity of substance
If residual coding is used to compress the image, then the code amount can be reduced, but the pipeline must be retraced when mode switching is required
Solution Approach 1:
The system determines the coding mode (residual or PCM) during the orthogonal transformation and quantization stage, which is before the variable-length coding stage. This preliminary decision allows the pipeline to proceed without interruption, avoiding the need to retrace previous stages when mode switching is needed.
Solution Approach 2:
The encoding process is segmented into distinct stages: orthogonal transformation/quantization (where mode decision is made) and variable-length coding (where actual coding occurs). This segmentation allows independent control of each stage, enabling mode switching without affecting the already-completed previous stages.
3Quantity of substance
If the maximum code value limit is strictly enforced in each block, then code amount control is achieved, but the encoding process may fail to complete in required time
Solution Approach 1:
By determining the coding mode during the orthogonal transformation and quantization stage (in advance), the system prepares all necessary processing results before reaching the variable-length coding stage. This eliminates the need for time-consuming pipeline retracing when code amount limits are approached, ensuring the encoding process completes within the required time.
Solution Approach 2:
The system changes the coding parameter (mode selection between residual and PCM) based on the code amount generated so far. When the code amount approaches the maximum limit, the system switches to PCM mode which uses simpler direct pixel value representation, reducing the time needed for variable-length coding while still meeting the code amount constraint.
Data Source
AI summary
A video image encoding device, in a first mode, variable-length-encodes a residual coefficient to generate a coefficient code string, outputs the coefficient code string and the header information in a state in which the header information is associated with the coefficient code string, in a second mode, directly uses a differential image as a coefficient code string without variable-length-encoding the differential image, and outputs the coefficient code string and the header information in a state in which the header information is associated with the coefficient code string.