Image Restoration Using Block-Based Motion Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image restoration methods for sequences, especially those with moving objects and backgrounds, face challenges in accurately utilizing temporal correlations due to high computational costs and noise in motion field estimations, leading to insufficient quality restoration and incompatibility with low-capacity systems.

Innovation Solution

A method that uses an external energy integrating global object movement to enhance segmentation, calculating internal energy derivatives for contour pixels and applying filtering to ensure regular movement variations, allowing for effective use of temporal correlations while reducing computational load.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If dense motion field estimation techniques are used to accurately capture temporal correlations between images, then image restoration quality is improved, but computational cost increases significantly

Engineering Contradiction:
Improveimage restoration qualityVSAvoidcomputational cost
Core Design Contradiction:
Measurement precisionVSPower

Solution Approach 1:

The patent divides the image into multiple blocks and processes each block independently using local motion estimation. This segmentation approach reduces the overall computational complexity while maintaining restoration quality by focusing computational resources on local regions rather than estimating dense motion fields for the entire image.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts and utilizes temporal correlations between successive images by comparing only relevant blocks. Instead of performing full dense motion field estimation, the method extracts essential motion information from temporal sequences to improve restoration quality with reduced computational overhead.

Inventive Principle:
Principle #2Taking out (Extraction)

2Productivity

If global motion estimation techniques are used to reduce computational cost, then processing efficiency is improved, but reliability of motion analysis deteriorates when scenes are non-stationary

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidmotion analysis reliability
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent applies local motion estimation to specific blocks rather than using a single global motion model for the entire image. This allows the system to adapt to local variations in motion patterns, improving reliability in non-stationary scenes while maintaining computational efficiency through localized processing.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent dynamically adapts the motion estimation approach by processing blocks independently and allowing different motion characteristics in different regions. This dynamic local adaptation enables the system to handle non-stationary scenes effectively without the high computational cost of global dense motion field estimation.

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If image segmentation is applied after motion analysis to differentiate object and background processing, then restoration effectiveness is improved, but device complexity increases

Engineering Contradiction:
Improverestoration effectivenessVSAvoidprocessing pipeline complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent integrates segmentation with motion analysis by dividing the image into blocks and applying motion estimation locally to each block. This approach effectively segments the processing task without requiring separate complex segmentation pipelines, thereby improving restoration effectiveness while controlling device complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent merges segmentation and motion analysis into a unified processing framework where block-based division serves both segmentation and motion estimation purposes. This combination reduces overall system complexity by eliminating the need for separate segmentation and motion analysis pipelines.

Inventive Principle:
Principle #5Merging (Combining)

Data Source

PatentEP3701492B1Image restoration method
Publication Date: 2022.03.02 SAFRAN ELECTRONICS & DEFENSE (FR)
  • EP3701492B1 patent drawingFigure 1~2
  • EP3701492B1 patent drawingFigure 3~4
  • EP3701492B1 patent drawingFigure 5~6

AI summary

The invention relates to a method for restoring images from a sequence of images, comprising, when applied to a first image from the image sequence: estimating (41) an information item representative of global motion of a background of the first image with respect to a second image; compensating (42) for the global background motion in the second image, so as to obtain a registered version of the second image, known as the registered second image; obtaining (43) a contour of an object in the first image by applying a segmentation method using the registered second image; using the contour of the object thus obtained to estimate (44) an information item representative of global motion of the object; and applying (45) an image restoration method to the first image using the estimated information items representative of the global background motion and of the global motion of the object.