Scalable Video Coding Upsampling and Bit-Depth Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In scalable video coding (SVC), performing upsampling and bit-depth conversion as separate processes can lead to reduced coding efficiency and prediction accuracy due to rounding errors, especially when both spatial and bit-depth scalabilities are present.
Innovation Solution
Combining upsampling and bit-depth conversion into a single-stage process to predict enhancement layer pixel information, using a single-stage process that contributes to both upsampling and bit-depth conversion, such as through a filter that performs simultaneous upsampling and bit-depth shifting, thereby reducing rounding errors and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If upsampling and bit-depth conversion are performed as separate processes, then each process can be optimized independently, but coding efficiency and prediction accuracy are reduced due to rounding errors
Solution Approach 1:
The patent combines upsampling and bit-depth conversion into a single integrated processing stage. The filter simultaneously performs upsampling operations and bit-depth conversion operations on the base layer pixel data, eliminating the need for separate sequential processing stages. This merging reduces the number of times rounding operations are performed, thereby reducing cumulative rounding errors and improving prediction accuracy in the enhancement layer.
Solution Approach 2:
The filter is designed to perform multiple functions simultaneously: it acts as both an upsampling filter and a bit-depth conversion filter. By making the filter universal, it can handle both spatial scaling (upsampling) and precision scaling (bit-depth conversion) in a single operation, reducing the overall complexity while maintaining or improving prediction accuracy.
2Productivity
If upsampling and bit-depth conversion are performed as separate processes, then implementation is simpler for each stage, but computational complexity increases and coding efficiency decreases
Solution Approach 1:
By merging upsampling and bit-depth conversion into a single filter operation, the patent reduces the total number of computational steps required. Instead of performing complete upsampling then complete bit-depth conversion in sequence, the integrated filter performs both operations concurrently on the same data, reducing redundant computations and improving overall coding efficiency.
Solution Approach 2:
The filter performs bit-depth conversion as a preliminary action during the upsampling process itself, rather than as a subsequent separate step. This preliminary conversion ensures that the data is prepared in the correct bit-depth format early in the processing chain, avoiding later reprocessing and improving overall computational efficiency.
3Measurement precision
If multiple separate processing stages are used for upsampling and bit-depth conversion, then each stage can be independently optimized, but prediction accuracy is reduced due to accumulated rounding errors
Solution Approach 1:
The patent merges multiple processing stages into a single filter operation that simultaneously performs upsampling and bit-depth conversion. This eliminates the sequential time required for separate processing stages and reduces the number of times rounding errors can accumulate, as there is only one set of rounding operations instead of multiple sequential ones.
Data Source
Figure 1
Figure 2A
Figure 2B
AI summary
An apparatus configured to code video information comprises a memory unit and a processor in communication with the memory unit. The memory unit is configured to store video information associated with a reference layer (RL) and an enhancement layer (EL). The EL comprises an EL video unit and the RL comprises an RL video unit corresponding to the EL video unit. The processor is configured to perform upsampling and bit-depth conversion on pixel information of the RL video unit in a single combined process to determine predicted pixel information of the EL video unit, and determine the EL video unit using the predicted pixel information.