Template-Based Filtering for Video Prediction Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video encoding and decoding techniques face challenges in achieving accurate predictions and efficient bitstream signaling, particularly in reducing residual data bits and minimizing signaling overhead, due to the lossy compression processes that introduce distortion and artifacts like blockiness.

Innovation Solution

The proposed solution involves applying filtering to prediction blocks before reconstruction, using template-based filtering to compare reference and current blocks, which reduces the need for signaling filter coefficients and improves the rate-distortion tradeoff by producing more accurate predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If filtering is applied to prediction blocks before reconstruction, then prediction accuracy is improved and residual data bits are reduced, but signaling overhead increases due to filter coefficients

Engineering Contradiction:
Improveprediction accuracyVSAvoidsignaling overhead
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The filter derives its coefficients self-service from the template data itself, comparing the reference block template with the current block template to automatically determine filtering parameters. This eliminates the need for external signaling of filter coefficients, as the filter uses the data it processes to configure itself.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

Template data serves as an intermediary that bridges the reference block and current block, enabling the filter to derive coefficients without direct signaling. The template comparison acts as a mediator that translates block differences into filtering parameters without requiring additional bitstream communication.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Productivity

If template-based filtering is used to compare reference and current blocks, then compression efficiency is improved by reducing residual data, but device complexity increases due to additional filtering operations

Engineering Contradiction:
Improvecompression efficiencyVSAvoidfiltering operation complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The filter performs preliminary actions by deriving coefficients from template comparison before the main reconstruction process. This preliminary filtering prepares the prediction block in advance, reducing the complexity of subsequent residual coding and reconstruction operations by pre-correcting prediction errors.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The filter dynamically changes its parameters (coefficients) based on the template comparison results. Instead of using fixed filtering parameters, the adapter adjusts filter characteristics adaptively according to the local block characteristics, improving compression efficiency while maintaining manageable complexity through parameter adaptation rather than structural complexity.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS20240236314A1Filtering applied to prediction in video coding
Publication Date: 2024.07.11 QUALCOMM INC
  • US20240236314A1 patent drawing
  • US20240236314A1 patent drawing
  • US20240236314A1 patent drawing

AI summary

A video decoder may be configured to determine a reference block for a current block of a current picture of video data, wherein the reference block is in the current picture; determine a prediction block based on the reference block; determine whether to apply filtering to the prediction block; based on determining that the filtering is to be applied to the prediction block, compare a template of the reference block to a template of the current block and filter the prediction block based on the comparing of the template of the reference block to the template of the current block to determine a filtered prediction block; decode the current block based on the filtered prediction block to determine a decoded version of the current block; and output a decoded picture of the video data comprising the decoded version of the current block.