Separable Bilinear Interpolation Filter for 3D Video Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current 3D-HEVC video coding techniques face inefficiencies due to high computational complexity and power consumption caused by non-separable filters, which result in increased clipping operations during advanced residual prediction (ARP) processes.

Innovation Solution

Implementing a separable bilinear interpolation filter to reduce the number of clipping operations by summing coefficients to 64, aligning with HEVC interpolation filters, and using this filter for ARP in both video encoding and decoding processes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If non-separable filters are used for advanced residual prediction in 3D-HEVC video coding, then coding precision is improved, but computational complexity and power consumption increase

Engineering Contradiction:
Improvecoding precisionVSAvoidcomputational complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The non-separable filter is divided into two separate one-dimensional filters (horizontal and vertical). Instead of applying a single two-dimensional non-separable filter directly, the patent applies the first one-dimensional filter along the horizontal direction and then applies the second one-dimensional filter along the vertical direction, segmenting the complex operation into simpler sequential steps that reduce computational complexity while maintaining filtering effectiveness

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent replaces the complex non-separable filtering mechanism with a simplified separable filtering mechanism. By substituting the mechanical operation of a complex non-separable filter with the sequential application of two simpler one-dimensional filters, the system achieves the same or similar filtering效果 with significantly reduced computational complexity and power consumption

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If non-separable filters are used for advanced residual prediction, then coding efficiency is improved, but the number of clipping operations increases

Engineering Contradiction:
Improvecoding efficiencyVSAvoidnumber of clipping operations
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The filtering operation is segmented into two separate one-dimensional filtering steps instead of one complex two-dimensional step. This segmentation reduces the number of clipping operations required, as each one-dimensional filter requires fewer clipping operations compared to a single non-separable filter, while still achieving the desired filtering effect for advanced residual prediction

Inventive Principle:
Principle #1Segmentation

3Measurement precision

If non-separable filters are used in ARP processes, then prediction accuracy is improved, but power consumption increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidpower consumption
Core Design Contradiction:
Measurement precisionVSUse of energy by stationary object

Solution Approach 1:

The patent substitutes the power-intensive non-separable filter mechanism with a more energy-efficient separable filter mechanism. By replacing the complex non-separable filtering operation with sequential one-dimensional filters, the system maintains prediction accuracy while significantly reducing power consumption, making the ARP process more suitable for energy-constrained devices

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP3162063B1Filters for advanced residual prediction in video coding
Publication Date: 2020.12.23 QUALCOMM INC
  • EP3162063B1 patent drawingFigure 1
  • EP3162063B1 patent drawingFigure 2
  • EP3162063B1 patent drawingFigure 3

AI summary

A video coder is configured to apply a separable bilinear interpolation filter when determining reference blocks as part of advanced residual prediction. Particularly, the video coder may determine, based on a motion vector of a current block in a current picture of video data, a location of a first reference block in a first reference picture. The video coder may also determine a location of a second reference block in a second reference picture. The video coder may apply a separable bilinear interpolation filter to samples of the second reference picture to determine samples of the second reference block. The video coder may apply the separable bilinear interpolation filter to samples of a third reference picture to determine samples of a third reference block. Each respective sample of a predictive block may be equal to a respective sample of the first reference block plus a respective residual predictor sample.