Adaptive Interpolation Filter Search for Lower-Complexity Video Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video compression techniques require inefficient and computationally expensive motion estimation and interpolation searches, especially for sub-pixel motion, leading to increased hardware size, complexity, and power consumption.
Innovation Solution
Adaptive interpolation filter search method that uses a single filter type to identify a closest-match motion vector, followed by calculating interpolation costs for additional filter types, reducing the number of calculations required and improving computational efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple filter types are used for sub-pixel motion estimation, then video quality is improved, but computational complexity and hardware requirements increase significantly
Solution Approach 1:
The patent divides the filter selection process into two stages: first identifying a closest-match motion vector using a single filter type, then performing interpolation cost calculations with additional filter types only for that specific vector. This segmentation reduces the overall computational burden while maintaining multiple filter type benefits.
Solution Approach 2:
The patent performs preliminary motion estimation using one filter type to identify the closest-match motion vector before conducting more expensive interpolation cost calculations with multiple filter types. This preliminary action eliminates the need to perform full multi-filter searches for all candidate vectors.
2Measurement precision
If multiple filter types are used for sub-pixel motion estimation, then video quality is improved, but power consumption increases
Solution Approach 1:
The patent segments the computationally intensive filtering operations into a preliminary single-filter stage and a subsequent multi-filter interpolation stage, reducing total power consumption by avoiding redundant calculations.
Solution Approach 2:
The patent applies partial action by using multiple filter types only partially - specifically for interpolation cost calculations on the closest-match vector - rather than applying them fully across all motion estimation operations, thereby reducing energy usage.
3Measurement precision
If exhaustive motion estimation search is performed, then encoding accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary motion estimation using a single filter type to quickly identify the closest-match motion vector, avoiding the need to perform exhaustive searches with multiple filter types from the beginning.
Solution Approach 2:
The patent performs partial exhaustive search by conducting interpolation cost calculations with multiple filter types only for the identified closest-match vector, rather than performing complete exhaustive searches for all candidate vectors.
Data Source
AI summary
Techniques for implementing adaptive interpolation filter search in video encoding are disclosed. Conventional interpolation filter searching is simplified to implement adaptive interpolation filter search by selecting one or more first filter types to determine one or more initial interpolation costs. After identifying an MV that produces a target interpolation error for one of the one or more first filter types, one or more secondary interpolation costs are calculated for one or more additional filter types based on the identified MV, and one of the one or more first filter type and one or more additional filter types that results in minimal interpolation error is selected as the interpolation filter type.