Fractional-Pixel Interpolation via Fixed and Adaptive Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression techniques, such as H.261 and MPEG, face inefficiencies in reducing temporal redundancy due to the high complexity and computational cost of fractional-pixel interpolation, especially when using adaptive filters, which require numerous multiplications and are not adaptable to varying video signals.
Innovation Solution
A method that combines fixed and adaptive filters for fractional-pixel interpolation, allowing for two-step interpolation with fewer multiplications and adaptable filter coefficients, reducing computational complexity while maintaining video signal quality, by using a fixed filter to generate intermediate subpixel images and an adaptive filter to produce reference subpixel images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If solely an adaptive filter is used for fractional-pixel interpolation, then video signal quality can be maintained, but computational complexity becomes very high requiring numerous multiplications
Solution Approach 1:
The patent segments the interpolation process into two distinct stages: a first interpolation stage using a fixed filter to generate intermediate subpixel values, and a second interpolation stage using an adaptive filter to generate final subpixel values. This segmentation reduces computational complexity by limiting the expensive adaptive filtering to only those pixels that truly require it, while using computationally efficient fixed filtering for the initial interpolation.
Solution Approach 2:
The patent applies different filter types to different spatial locations based on local image characteristics. The adaptive filter is selectively applied to pixels located at motion boundaries or regions with high gradient values where precision is critical, while fixed filters are used in regions with uniform characteristics where high precision is less critical. This local differentiation maintains video signal quality at critical locations while reducing overall computational complexity.
2Device complexity
If solely a fixed filter is used for interpolation, then computational complexity is reduced, but the filter cannot capture small motion offsets and is not adaptable to varying video signals
Solution Approach 1:
The patent introduces dynamic adaptability into the interpolation system by using an adaptive filter in the second stage that can adjust its coefficients based on local image characteristics such as gradient values and motion boundary detection. This dynamic adaptation allows the system to respond to varying video signal characteristics while maintaining reasonable computational complexity through the two-stage architecture.
Solution Approach 2:
The fixed filter acts as an intermediary that prepares intermediate subpixel values which then serve as input to the adaptive filter. This intermediary step allows the system to combine the computational efficiency of fixed filtering with the adaptability of adaptive filtering, as the fixed filter pre-processes the image data to reduce the burden on the subsequent adaptive filtering stage.
3Measurement precision
If a 36-tap adaptive filter is used to interpolate each subpixel, then interpolation quality improves, but each subpixel requires 36 multiplications increasing computational cost
Solution Approach 1:
The patent applies partial adaptive filtering by using a reduced number of taps in the adaptive filter compared to a full 36-tap filter. By combining the fixed filter's contribution with a smaller adaptive filter, the system achieves sufficient interpolation quality without requiring the full computational power of a 36-tap adaptive filter for every pixel, thus reducing the number of multiplications required.
Data Source
AI summary
A method for use in video compression is disclosed. In particular, the claimed invention relates to a method of more efficient fractional-pixel interpolation in two steps by a fixed filter (240) and an adaptive filter (250) for fractional-pixel motion compensation.