Context-Adaptive Interpolation Filters for Fractional-Motion Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently generating prediction samples for fractional motion vectors or block vectors, particularly in inter prediction, due to the lack of adaptive interpolation filters that can effectively minimize distortion and adapt to varying video content statistics.
Innovation Solution
The method involves deriving context-adaptive interpolation filters using reconstructed samples from neighboring templates of a current and reference block, applying these filters to generate prediction samples by minimizing distortion, and selecting the optimal filter based on quantified metrics such as SAD, SSD, or SSE.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If fixed interpolation filters are used for fractional motion vectors, then device complexity is reduced, but prediction accuracy deteriorates due to inability to adapt to varying video content statistics
Solution Approach 1:
The patent implements dynamic interpolation filters that adapt to local video content characteristics. The filter derivation process uses reconstructed samples from current and reference blocks to compute context-adaptive filters, allowing the system to dynamically adjust filter coefficients based on local statistics rather than using fixed filters throughout.
Solution Approach 2:
The patent changes the parameters of interpolation filters based on local video content statistics. By deriving filter coefficients from actual reconstructed samples in the vicinity of current and reference blocks, the system adapts filter parameters to match local image characteristics, thereby improving prediction accuracy without requiring overly complex external control mechanisms.
2Measurement precision
If context-adaptive interpolation filters are derived using reconstructed samples, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent performs preliminary actions by using already-reconstructed samples from the decoding process to derive interpolation filters. Rather than requiring additional expensive computations, the system leverages samples that are already available in the reconstruction buffer, minimizing additional computational overhead while achieving adaptive filtering.
Solution Approach 2:
The system uses its own reconstructed samples to generate the interpolation filters it needs. The decoded video data serves dual purposes: both as the output to be improved and as the source material for deriving the filtering parameters, creating a self-sufficient system that avoids external computational dependencies.
3Measurement precision
If adaptive interpolation filters are applied, then distortion is reduced, but processing time increases
Solution Approach 1:
The patent applies adaptive filtering selectively rather than uniformly across all blocks. By deriving context-adaptive filters only where needed based on local content characteristics, the system achieves distortion minimization in critical areas while avoiding unnecessary processing time expenditure in regions where fixed filters suffice.
Data Source
AI summary
This disclosure relates generally to video coding and particularly to methods and systems for deriving context adaptive interpolation filter used for generating prediction samples in inter prediction involving a fractional motion vector or block vector. For example, an encoder and a decoder may derive interpolation filters adaptively using reconstructed samples. The reconstructed samples may belong to templates near a current block and a reference block related by the motion vector or block vector. The derivation of the adaptive interpolation filters may be based on minimizing distortions between interpolated templates associated with the reference block and the templates associated with the current block.