Subpixel 4-Tap Interpolation Filtering for Low-Memory Video Prediction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in achieving high compression ratios with minimal sacrifice in picture quality, particularly in streaming and storage applications where bandwidth and memory resources are limited.

Innovation Solution

The method involves using a subpixel 4-tap interpolation filter with specific filter coefficients defined by the equations c0=1⁢6-p2, c1=1⁢6+1⁢6-p2, c2=1⁢6+p2, and c3=p2, which are calculated analytically based on the subpixel offset value, allowing for efficient intra- and inter-prediction processing in video frames.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of energy

If conventional video compression techniques are used, then bandwidth usage is reduced, but picture quality deteriorates

Engineering Contradiction:

Improvebandwidth usageVSAvoidpicture quality

Core Design Contradiction:

Loss of energyVSManufacturing precision

Solution Approach 1:

The patent applies parameter changes by using analytical filter coefficients that adapt to subpixel offset values. The filter coefficients are calculated based on the fractional part of the subpixel offset, allowing the system to maintain high prediction accuracy across different motion scenarios. This adaptive parameter adjustment enables better picture quality at lower bitrates by optimizing the interpolation process for each specific motion case.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces traditional lookup table-based filter coefficient selection with analytical calculations. Instead of storing and searching through pre-computed coefficient tables, the system directly computes filter coefficients using mathematical formulas based on subpixel offset values. This substitution reduces memory access requirements and enables more precise coefficient selection, improving both compression efficiency and picture quality.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If complex interpolation filters are used, then prediction accuracy is improved, but device complexity increases

Engineering Contradiction:

Improveprediction accuracyVSAvoiddevice complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex lookup table mechanisms with simple analytical calculations. The filter coefficients are computed directly using mathematical formulas involving basic arithmetic operations on subpixel offset values. This substitution maintains high prediction accuracy while significantly reducing device complexity by eliminating the need for large coefficient tables and complex selection logic.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent uses parameter changes by deriving filter coefficients analytically from subpixel offset parameters. The fractional part of the subpixel offset directly determines the filter coefficients through simple mathematical relationships. This approach achieves high prediction accuracy with minimal computational complexity, as the parameters naturally encode the required filter characteristics without requiring complex processing.

Inventive Principle:

Principle #35Parameter changes

3Speed

If lookup tables for filter coefficients are stored, then filtering speed is improved, but memory demands increase

Engineering Contradiction:

Improvefiltering speedVSAvoidmemory demands

Core Design Contradiction:

SpeedVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential information needed for filter coefficient determination from complex lookup tables. By using analytical formulas that compute coefficients directly from subpixel offset values, the system extracts the core functional relationship without requiring storage of extensive coefficient tables. This extraction eliminates unnecessary memory storage while maintaining filtering speed through direct calculation.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent substitutes memory-based coefficient retrieval with calculation-based coefficient generation. Instead of reading coefficients from stored tables, the system computes them on-the-fly using analytical formulas. This substitution eliminates memory access operations entirely, reducing both memory demands and potential bottlenecks, while the simplicity of the formulas ensures computation remains fast and efficient.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12348788B2Method and apparatus of interpolation filtering for predictive coding

Publication Date: 2025.07.01 HUAWEI TECH CO LTD

AI summary

The present disclosure relates to intra- or inter-prediction for video encoding and decoding. For that purpose, an apparatus and methods obtain a reference sample and a subpixel offset value. A 4-tap interpolation filter is used to filter the reference sample to obtain a predicted sample value. The filter coefficients c0, c1, c2, and c3 of the 4-tap interpolation filter satisfyc0=1⁢6-p2,c1=1⁢6+1⁢6-p2,c2=1⁢6+p2,andc3=p2,with p being a fractional part of the value of the subpixel offset.