Motion Compensation of High-Frequency Image Detail to Reduce Flicker

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Upscaling low-resolution images to high-resolution images using neural networks often results in inconsistent added texture across frames, leading to visible flicker and a deteriorated user experience.

Innovation Solution

A system that includes an image generator using a neural network to enhance images and a motion analyzer to perform motion compensation on high-frequency components, adjusting them based on motion vectors to generate an adjusted enhanced image, thereby reducing flicker by selectively applying temporal filtering.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If neural network upscaling is applied to low-resolution images, then image resolution and texture quality are improved, but temporal consistency deteriorates causing visible flicker

Engineering Contradiction:
Improveimage resolutionVSAvoidtemporal consistency
Core Design Contradiction:
Manufacturing precisionVSStability of the object's composition

Solution Approach 1:

The image is divided into frequency components (low-frequency and high-frequency parts). Motion compensation is selectively applied only to the high-frequency component that causes flicker, while the low-frequency component remains unchanged. This segmentation allows targeted correction of temporal inconsistency without affecting overall image quality.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different processing approaches are applied to different parts of the image based on their frequency characteristics. The high-frequency component (which contains texture details and is prone to flicker) receives motion compensation processing, while the low-frequency component (which contains overall structure) is preserved as-is. This local quality approach ensures that correction is applied where needed without unnecessary processing elsewhere.

Inventive Principle:
Principle #3Local quality

2Stability of the object's composition

If temporal filtering is applied to reduce flicker, then temporal consistency is improved, but image texture quality deteriorates

Engineering Contradiction:
Improvetemporal consistencyVSAvoidimage texture quality
Core Design Contradiction:
Stability of the object's compositionVSManufacturing precision

Solution Approach 1:

The image processing is segmented by frequency components. Motion compensation is applied exclusively to the high-frequency component that exhibits flicker, while the low-frequency component is left unprocessed. This selective segmentation prevents the degradation of overall image quality that would result from applying temporal filtering to the entire image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different quality preservation strategies are applied to different frequency domains. The high-frequency component undergoes motion-based correction to maintain temporal consistency, while the low-frequency component maintains its original texture quality. This local quality approach ensures that texture degradation is avoided in regions where it is not necessary.

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP4374577B1Motion compensation for neural network enhanced images
Publication Date: 2026.04.22 QUALCOMM INC
  • EP4374577B1 patent drawingFigure 1
  • EP4374577B1 patent drawingFigure 2A~2B
  • EP4374577B1 patent drawingFigure 2C

AI summary

A device includes a memory and one or more processors. The memory is configured to store instructions. The one or more processors are configured to execute the instructions to apply a neural network to a first image to generate an enhanced image. The one or more processors are also configured to execute the instructions to adjust at least a portion of a high-frequency component of the enhanced image based on a motion compensation operation to generate an adjusted high-frequency image component. The one or more processors are further configured to execute the instructions to combine a low-frequency component of the enhanced image and the adjusted high-frequency image component to generate an adjusted enhanced image.