Cubic Spline Filter Design for Non-Dyadic Video Upsampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current Scalable Video Coding (SVC) extension of H.264/AVC is limited in its ability to handle non-dyadic spatial scalability, such as SD to HD scalability, as it only addresses configurations where the ratio of picture width to height between successive layers is 2, restricting its application in broadcasting and other general scenarios.
Innovation Solution
The development of matrix-based representations for 6-tap and 4-tap cubic spline filters, which offer similar or improved frequency responses compared to existing filters like Lanczos-3 and Catmull-Rom, allowing for adaptive upsampling with adjustable weighting factors and reduced complexity, enabling sharper image interpolation and wider pass-bands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 6-tap Lanczos-3 filters are used for upsampling, then frequency response accuracy is improved, but device complexity increases
Solution Approach 1:
The patent creates a simplified copy of the Lanczos-3 filter by deriving equivalent 4-tap and 6-tap cubic spline filters that replicate the frequency response characteristics without requiring the full complexity of the original Lanczos-3 implementation. This allows the system to achieve similar frequency response accuracy with reduced computational burden.
Solution Approach 2:
The patent transforms the filter design by changing the mathematical parameters from Lanczos-3 sinc-function-based coefficients to cubic spline polynomial coefficients. This parameter transformation maintains the essential frequency response properties while enabling more efficient computation and simpler implementation structures.
2Device complexity
If 4-tap Catmull-Rom filters are used for upsampling, then device complexity is reduced, but pass-band width is limited
Solution Approach 1:
The patent introduces dynamic adaptability by providing multiple filter designs (4-tap and 6-tap cubic spline filters) that can be selected based on the specific application requirements. The system can dynamically choose between different filter lengths and characteristics to optimize both complexity and pass-band width for different scenarios.
Solution Approach 2:
The patent combines elements of different filter designs by providing a family of cubic spline filters with varying tap lengths (4-tap and 6-tap versions). This composite approach allows the system to achieve both low complexity (using 4-tap) and wide pass-band (using 6-tap) by selecting the appropriate filter from the family based on specific needs.
3Productivity
If standard upsampling filters are used in SVC, then coding efficiency is maintained, but adaptability to non-dyadic scalability is limited
Solution Approach 1:
The patent creates universal cubic spline filter designs that can function across multiple scalability scenarios including both dyadic and non-dyadic spatial scalability. The filters are designed to be broadly applicable to different video coding contexts while maintaining coding efficiency, thus achieving multi-functionality that addresses both adaptability and productivity requirements.
Data Source
AI summary
Aspects of the present invention relate to systems, methods and devices for upsampling images and design of upsampling filters. Some aspects relate to a determination of a phase offset position in a higher resolution picture relative to a lower resolution picture. Interpolation filter coefficients for some filters may then be selected based on the filter offset. Other aspects relate to selection of coefficients for filters that are not dependent on the phase offset. In certain implementations, a weighting factor may be used to combine the effects of a phase-offset-dependent filter and an independent filter.