Multi-Stage Motion Compensation for HDR Media Ghosting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for generating High Dynamic Range (HDR) media face challenges in handling noise, halos, dynamic range, and ghosting, especially in saturated and dark regions, and require extensive IQ tuning and struggle with false ghosts due to improper exposure alignment and image registration.

Innovation Solution

The proposed solution involves a multi-stage compensation of motion in HDR media processing, using a processing module with exposure alignment, ghost map generation, and weight map calculation to correct exposure and registration errors, thereby minimizing ghosting and enhancing dynamic range in HDR images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Illumination intensity

If multiple image frames of different exposures are captured and blended together to form HDR image, then dynamic range is improved, but ghosting and halos increase

Engineering Contradiction:
Improvedynamic rangeVSAvoidghosting and halos
Core Design Contradiction:
Illumination intensityVSObject-generated harmful factors

Solution Approach 1:

The patent segments the ghost removal process into multiple stages: first identifying ghost regions through photometric difference maps, then separately processing different types of ghosts (motion-induced vs. misalignment-induced), and finally applying targeted corrections to specific regions rather than processing the entire image uniformly. This segmentation allows selective removal of ghosts while preserving legitimate HDR content.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by first aligning exposures and registering images before blending, and by pre-identifying ghost regions using photometric difference maps. This preliminary detection and alignment prevents ghosting from occurring in the final HDR composite, rather than attempting to correct it afterward.

Inventive Principle:
Principle #10Preliminary action

2Illumination intensity

If the number of image frames for blending is increased to improve dynamic range and reduce noise, then extreme ghosting increases

Engineering Contradiction:
Improvedynamic range and noiseVSAvoidextreme ghosting
Core Design Contradiction:
Illumination intensityVSObject-generated harmful factors

Solution Approach 1:

The patent implements feedback mechanisms by using photometric difference maps to detect ghost regions, then feeding this information back into the blending process to adjust weights and apply corrections. The system continuously monitors for ghosting artifacts and dynamically adjusts processing parameters based on detected ghost severity and type.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent applies local quality by treating different regions of the image differently - applying aggressive ghost removal only in regions identified as containing ghosts, while preserving original HDR content in clean regions. This allows the system to handle extreme ghosting locally without degrading the overall image quality or losing legitimate high-dynamic-range information.

Inventive Principle:
Principle #3Local quality

3Productivity

If conventional exposure alignment method is used, then processing speed is maintained, but false ghosts increase in bright and dark regions

Engineering Contradiction:
Improveprocessing speedVSAvoidfalse ghosts
Core Design Contradiction:
ProductivityVSObject-generated harmful factors

Solution Approach 1:

The patent employs dynamic processing by adapting the exposure alignment method based on local image characteristics. In bright and dark regions where false ghosts are likely, the system dynamically switches to more robust alignment techniques or applies corrective filters, while maintaining faster processing in uniform regions. This dynamic adaptation reduces false ghosts without significantly impacting overall processing speed.

Inventive Principle:
Principle #15Dynamics

4Object-generated harmful factors

If multi exposure fusion is not rightly handled, then artefacts such as halos and ghosts increase, but dynamic range and noise performance are sacrificed

Engineering Contradiction:
ImproveartefactsVSAvoiddynamic range and noise
Core Design Contradiction:
Object-generated harmful factorsVSIllumination intensity

Solution Approach 1:

The patent introduces intermediary processing steps including photometric difference maps as intermediaries to detect ghost regions, and weight maps as intermediaries to control the blending process. These intermediaries mediate between the raw multi-exposure frames and the final HDR output, allowing sophisticated ghost removal while preserving the dynamic range and noise characteristics of the original frames.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3834170B1Apparatus and methods for generating high dynamic range media, based on multi-stage compensation of motion
Publication Date: 2024.06.19 SAMSUNG ELECTRONICS CO LTD
  • EP3834170B1 patent drawingFigure 1a
  • EP3834170B1 patent drawingFigure 1b
  • EP3834170B1 patent drawingFigure 2~3

AI summary

Apparatus and methods for generating High Dynamic Range (HDR) media, based on multi-stage compensation of motion in a captured scene is disclosed in the embodiments herein. Embodiments herein relates to the field of image processing devices suitable for processing two or more images of different exposures and more particularly to apparatus and methods for generating a High Dynamic Range (HDR) media, based on a multi-stage compensation of motion in a captured scene. The apparatus is configured to correct exposure alignment error and media registration error from a registered plurality of media frames comprising a plurality of exposure levels corresponding to the captured scene. The apparatus is configured to remove a plurality of false ghost artefacts in the plurality of media frames. The apparatus is configured to generate the HDR media, based on a generated ghost map.