Optical-Flow Image Fusion for Mixed-Motion Video Denoising

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image processing methods struggle to accurately calculate motion vector information between adjacent frames when global motion of the image capturing device and local motion of moving objects occur simultaneously, leading to poor alignment and issues like ghosting, noising, or blurring during video multi-frame de-noising.

Innovation Solution

An image processing method that utilizes feature information from regions identified through motion estimation based on optical flow to separate local and global motion regions, applying different parameters for accurate fusion and alignment of frames, thereby improving inter-frame alignment and reducing ghosting.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If motion estimation is performed between adjacent frames to enable multi-frame de-noising, then de-noising capability is improved, but motion detection accuracy deteriorates when global motion and local motion occur simultaneously

Engineering Contradiction:
Improvede-noising capabilityVSAvoidmotion detection accuracy
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent segments the image frame into multiple regions (first region, second region, third region, fourth region) based on motion characteristics. By dividing the image into different regions with distinct motion properties, the system can apply region-specific processing to accurately detect both global motion and local motion separately, resolving the contradiction between maintaining de-noising capability and improving motion detection accuracy.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies different processing parameters and quality standards to different regions of the image. The first region (with optical flow motion estimation) and second region (with pixel association conditions) are treated differently, allowing the system to optimize motion detection for each region's specific characteristics while maintaining overall de-noising effectiveness.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If feature information from multiple regions is used for frame fusion, then inter-frame alignment accuracy is improved, but processing complexity increases

Engineering Contradiction:
Improveinter-frame alignment accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the image into multiple regions and processes each region separately with appropriate feature extraction methods. This segmentation allows the system to manage complexity by handling smaller, more manageable region-specific tasks rather than processing the entire image uniformly, while still achieving high overall alignment accuracy through the integration of regional results.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4105886B1Image processing method and device
Publication Date: 2025.10.08 GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
  • EP4105886B1 patent drawingFigure 1~2
  • EP4105886B1 patent drawingFigure 3~4
  • EP4105886B1 patent drawingFigure 5~6

AI summary

Disclosed in the embodiments of the present application are an image processing method and apparatus, and a device. The method comprises: obtaining the feature information of a first region in the current image frame, wherein the first region comprises a region that is determined in the current image frame by performing, on the current image frame and a previous image frame, motion estimation based on an optical flow method; obtaining the feature information of a second region in the current image frame, wherein the second region comprises a region corresponding to pixel points having an association relationship, meeting a condition, between a pixel point of a plurality of first pixel points of the current image frame and a pixel point of a plurality of second pixel points of the previous image frame; and based on the feature information of the first region and that of the second region, fusing the previous image frame and the current image frame to obtain the processed current image frame, wherein the processed current image frame is used as a previous image frame of a next image frame for processing the next image frame.