SlideShare a Scribd company logo
1 of 21
Color transfer
between high-
dynamic-range
images
H. Hristova, R. Cozot,
O. Le Meur, K. Bouatouch
University of Rennes 1
Rennes, France
Outline
● Introduction
- Main objective
- Contributions
● Extension to the HDR domain of a color
transfer method
● Results and evaluation
● Generalization for state-of-the-art color
transfer methods
● Conclusion
2
Main goal
● Carrying out a color transfer between two HDR images
directly in the HDR domain
Input Reference
3
● Solution: apply color transfer methods to stylize an HDR
image with regards to a reference image
Why do LDR color transfer methods need to
be extended to the HDR domain?
● LDR color spaces
- well predict the color gamut for luminance levels
between zero and the display white point
- uncertain applicability to HDR images
● Color trend above the perfect diffuse white
4
Why do LDR color transfer methods need to
be extended to the HDR domain?
● Assumption: a unique multivariate Gaussian distribution
● HDR domain: to fit the high range of lightness of HDR
images we need to assume mixture of Gaussian
distributions
5
Why do LDR color transfer methods need to
be extended to the HDR domain?
● Lightness - approximated by luminance in the LDR
domain
● HDR domain - distinguish between the absolute
luminance and the lightness (the L channel of CIE Lab)
6
Contributions
● Adaptation of [Hristova et al., 2015] color
transfer method to HDR images
- HDR color spaces
- Modifications of the clustering step and of the
image classification
● Cluster-based local chromatic adaptation
transform
● Generalization for state-of-the-art color transfer
methods
7
8
Extension to HDR images
• Linear search for significant peaks in the image hue histogram
- Colors-based style images: more than one significant color cluster
- Light-based style images: one significant color cluster
Input and
reference
images
Color
space
conversion
Image
classification
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
[Hristovaetal.,2015]
• The number of significant peaks determines the number of clusters
- Colors-based style images: hue histogram
- Light-based style images: luminance histogram
Extension to HDR images
Input and
reference
images
Color
space
conversion
Image
classification
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
9
[Hristovaetal.,2015]
LDR
images
CIE Lab
L channel of
CIE Lab
HDR
images
hdr-CIELab
Log-
luminance
● Dashed line: cubic function of L channel
(CIE Lab)
● Solid line: Michaelis-Menten function by
which we replace the cubic function of L
channel (CIE Lab)
● hdr-CIELab color space [Fairchild et al.,
2004]
[Fairchild et al., 2004]
ModificationsModified
Extension to HDR images
10
LDR
images
CIE Lab
L channel of
CIE Lab
L-based
clustering
HDR
images
hdr-CIELab
Log-
luminance
Log-
luminance
clustering
[Hristovaetal.,2015]
Logarithmic
transform
Input and
reference
images
Color
space
conversion
Image
classification
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
ModifiedModifications
Extension to HDR images
11
Local CAT
Cluster-
based local
CAT
[Hristovaetal.,2015]
LDR
images
CIE Lab
L channel of
CIE Lab
L-based
clustering
HDR
images
hdr-CIELab
Log-
luminance
Log-
luminance
clustering
Input and
reference
images
Color
space
conversion
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
Image
classification
ModifiedModifications
Extension to HDR images
12
Gaussianlow-passfilter
(h)
(m)
(sh)
(sh) (m) (h)
Local CAT
Cluster-
based local
CAT
[Hristovaetal.,2015]
LDR
images
CIE Lab
L channel of
CIE Lab
L-based
clustering
HDR
images
hdr-CIELab
Log-
luminance
Log-
luminance
clustering
Input and
reference
images
Color
space
conversion
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
Image
classification
ModifiedModifications
Extension to HDR images
13
Input ReferenceCluster-based local CAT
Local CAT
Cluster-
based local
CAT
[Hristovaetal.,2015]
LDR
images
CIE Lab
L channel of
CIE Lab
L-based
clustering
HDR
images
hdr-CIELab
Log-
luminance
Log-
luminance
clustering
Input and
reference
images
Color
space
conversion
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
Image
classification
ModifiedModifications
Extension to HDR images
14
Input ReferenceCluster-based local CAT
Local CAT
Cluster-
based local
CAT
[Hristovaetal.,2015]
LDR
images
CIE Lab
L channel of
CIE Lab
L-based
clustering
HDR
images
hdr-CIELab
Log-
luminance
Log-
luminance
clustering
Input and
reference
images
Color
space
conversion
Clustering
and
mapping
Color
transfer
Chromatic
adaptation
transform
Final result
Image
classification
ModifiedModifications
Objective evaluation of the results
15
● 10 image pairs
● Two tone-mapping operators: [Durand et al., 2002] and
[Reinhard et al., 2002]
● SSIM and Bhattacharya coefficient
Results
16
Input
Reference
Color transfer with CAT
Color transfer without CAT
Color transfer with CAT
Color transfer without CAT
[Hristova et al., 2015] HDR extension
Generalization and results
17
[Reinhard et al., 2001] - global method [Tai et al., 2005] - clustering (local
transformations)
Input
Reference
Generalization and results
18
[Pitié et al., 2007] - CIE Lab [Pitié et al., 2007] - hdr-CIELab
Input
Reference
Generalization and results
19
[Bonneel et al., 2013] - luminance clustering [Bonneel et al., 2013] - log-luminance
clustering
Input
Reference
Conclusion
● Extension of a novel local color transfer
method [Hristova et al., 2015]
- Modifications to CIE Lab -> hdr-CIELab
- Luminance/Lightness -> Log-luminance
● Generalization to state-of-the-art methods
● Future work
- Need for a more precise color
mapping/color transformation between two
HDR images
- Need for better HDR color spaces
20
Thank you for your attention!
21

More Related Content

Viewers also liked

Make the Most of Your Business Travels: Things to do in Dallas, TX
Make the Most of Your Business Travels: Things to do in Dallas, TXMake the Most of Your Business Travels: Things to do in Dallas, TX
Make the Most of Your Business Travels: Things to do in Dallas, TXKing of Maids
 
Dicas de viagem dubai sokan
Dicas de viagem dubai sokanDicas de viagem dubai sokan
Dicas de viagem dubai sokanchinaturismo
 
бессмертный полк в Туле
бессмертный полк в Тулебессмертный полк в Туле
бессмертный полк в ТулеAlexander Shneiderman
 
How Makerere university is idealy thriving on her namethat she already made l...
How Makerere university is idealy thriving on her namethat she already made l...How Makerere university is idealy thriving on her namethat she already made l...
How Makerere university is idealy thriving on her namethat she already made l...WALIASH
 
Importancia de la educacion...
Importancia de la educacion...Importancia de la educacion...
Importancia de la educacion...jaimejuly
 
Dockercon EU 2015 Recap
Dockercon EU 2015 RecapDockercon EU 2015 Recap
Dockercon EU 2015 RecapLee Calcote
 
Methods for comparing scanpaths and saliency maps: strengths and weaknesses
Methods for comparing scanpaths and saliency maps: strengths and weaknessesMethods for comparing scanpaths and saliency maps: strengths and weaknesses
Methods for comparing scanpaths and saliency maps: strengths and weaknessesOlivier Le Meur
 
Capitulo 10 10 09_2008_11_54_42
Capitulo 10 10 09_2008_11_54_42Capitulo 10 10 09_2008_11_54_42
Capitulo 10 10 09_2008_11_54_42carolina andrea
 
The Events department
The Events departmentThe Events department
The Events departmentOscar Barraza
 
Leadership at IBM - Report on Pat O'Sullivan
Leadership at IBM - Report on Pat O'SullivanLeadership at IBM - Report on Pat O'Sullivan
Leadership at IBM - Report on Pat O'SullivanPat O'Sullivan
 
E-safety Impero slides Mar2015
E-safety Impero slides Mar2015E-safety Impero slides Mar2015
E-safety Impero slides Mar2015James Grew
 

Viewers also liked (20)

Vsgames2010 v3
Vsgames2010 v3Vsgames2010 v3
Vsgames2010 v3
 
Make the Most of Your Business Travels: Things to do in Dallas, TX
Make the Most of Your Business Travels: Things to do in Dallas, TXMake the Most of Your Business Travels: Things to do in Dallas, TX
Make the Most of Your Business Travels: Things to do in Dallas, TX
 
Tondoon
TondoonTondoon
Tondoon
 
Dicas de viagem dubai sokan
Dicas de viagem dubai sokanDicas de viagem dubai sokan
Dicas de viagem dubai sokan
 
бессмертный полк в Туле
бессмертный полк в Тулебессмертный полк в Туле
бессмертный полк в Туле
 
How Makerere university is idealy thriving on her namethat she already made l...
How Makerere university is idealy thriving on her namethat she already made l...How Makerere university is idealy thriving on her namethat she already made l...
How Makerere university is idealy thriving on her namethat she already made l...
 
Artful Pools Design and Consulting
Artful Pools Design and ConsultingArtful Pools Design and Consulting
Artful Pools Design and Consulting
 
Importancia de la educacion...
Importancia de la educacion...Importancia de la educacion...
Importancia de la educacion...
 
Hijrah Nabi
Hijrah NabiHijrah Nabi
Hijrah Nabi
 
Dockercon EU 2015 Recap
Dockercon EU 2015 RecapDockercon EU 2015 Recap
Dockercon EU 2015 Recap
 
Methods for comparing scanpaths and saliency maps: strengths and weaknesses
Methods for comparing scanpaths and saliency maps: strengths and weaknessesMethods for comparing scanpaths and saliency maps: strengths and weaknesses
Methods for comparing scanpaths and saliency maps: strengths and weaknesses
 
Capitulo 10 10 09_2008_11_54_42
Capitulo 10 10 09_2008_11_54_42Capitulo 10 10 09_2008_11_54_42
Capitulo 10 10 09_2008_11_54_42
 
Sbindl
SbindlSbindl
Sbindl
 
Kronologi
KronologiKronologi
Kronologi
 
The Events department
The Events departmentThe Events department
The Events department
 
Embedded Systems Report
Embedded Systems ReportEmbedded Systems Report
Embedded Systems Report
 
JMS Innovation
JMS InnovationJMS Innovation
JMS Innovation
 
Leadership at IBM - Report on Pat O'Sullivan
Leadership at IBM - Report on Pat O'SullivanLeadership at IBM - Report on Pat O'Sullivan
Leadership at IBM - Report on Pat O'Sullivan
 
Les Journées de la Francophonie 2014
Les Journées de la Francophonie 2014Les Journées de la Francophonie 2014
Les Journées de la Francophonie 2014
 
E-safety Impero slides Mar2015
E-safety Impero slides Mar2015E-safety Impero slides Mar2015
E-safety Impero slides Mar2015
 

Similar to Color transfer between high-dynamic-range images

presentation644v4
presentation644v4presentation644v4
presentation644v4Maikon
 
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...Sumadeep Juvvalapalem
 
Deep Photo Style Transfer from Adobe
Deep Photo Style Transfer from AdobeDeep Photo Style Transfer from Adobe
Deep Photo Style Transfer from AdobeJhe-Wei Lee
 
digital image processing color processing
digital image processing color processingdigital image processing color processing
digital image processing color processingrajaramsharath
 
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processing
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processingDIGITAL SIGNAL PROCESSING - Day 3 colour Image processing
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processingvijayanand Kandaswamy
 
HDR Insights Article 3: Understanding HDR Tone Mapping
HDR Insights Article 3: Understanding HDR Tone MappingHDR Insights Article 3: Understanding HDR Tone Mapping
HDR Insights Article 3: Understanding HDR Tone MappingVeneraTech
 
HDR Insights Article 2 : PQ and HLG transfer functions for HDR
HDR Insights Article 2 : PQ and HLG transfer functions for HDRHDR Insights Article 2 : PQ and HLG transfer functions for HDR
HDR Insights Article 2 : PQ and HLG transfer functions for HDRVeneraTech
 
A new Post-Processing Pipeline
A new Post-Processing PipelineA new Post-Processing Pipeline
A new Post-Processing PipelineWolfgang Engel
 
A modern Post-Processing Pipeline
A modern Post-Processing PipelineA modern Post-Processing Pipeline
A modern Post-Processing PipelineWolfgang Engel
 
High Dynamic Range Imaging- A Review
High Dynamic Range Imaging- A ReviewHigh Dynamic Range Imaging- A Review
High Dynamic Range Imaging- A ReviewCSCJournals
 

Similar to Color transfer between high-dynamic-range images (12)

presentation644v4
presentation644v4presentation644v4
presentation644v4
 
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...
COMPARISION OF TONE-MAPPING ALGORITHM BASED ON STRUCTURAL FIDELITY AND STATIS...
 
Deep Photo Style Transfer from Adobe
Deep Photo Style Transfer from AdobeDeep Photo Style Transfer from Adobe
Deep Photo Style Transfer from Adobe
 
HDR and WCG Principles-Part 6
HDR and WCG Principles-Part 6HDR and WCG Principles-Part 6
HDR and WCG Principles-Part 6
 
Color Image Processing
Color Image ProcessingColor Image Processing
Color Image Processing
 
digital image processing color processing
digital image processing color processingdigital image processing color processing
digital image processing color processing
 
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processing
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processingDIGITAL SIGNAL PROCESSING - Day 3 colour Image processing
DIGITAL SIGNAL PROCESSING - Day 3 colour Image processing
 
HDR Insights Article 3: Understanding HDR Tone Mapping
HDR Insights Article 3: Understanding HDR Tone MappingHDR Insights Article 3: Understanding HDR Tone Mapping
HDR Insights Article 3: Understanding HDR Tone Mapping
 
HDR Insights Article 2 : PQ and HLG transfer functions for HDR
HDR Insights Article 2 : PQ and HLG transfer functions for HDRHDR Insights Article 2 : PQ and HLG transfer functions for HDR
HDR Insights Article 2 : PQ and HLG transfer functions for HDR
 
A new Post-Processing Pipeline
A new Post-Processing PipelineA new Post-Processing Pipeline
A new Post-Processing Pipeline
 
A modern Post-Processing Pipeline
A modern Post-Processing PipelineA modern Post-Processing Pipeline
A modern Post-Processing Pipeline
 
High Dynamic Range Imaging- A Review
High Dynamic Range Imaging- A ReviewHigh Dynamic Range Imaging- A Review
High Dynamic Range Imaging- A Review
 

More from Olivier Le Meur

Guided tour of visual attention
Guided tour of visual attentionGuided tour of visual attention
Guided tour of visual attentionOlivier Le Meur
 
Your gaze betrays your age
Your gaze betrays your ageYour gaze betrays your age
Your gaze betrays your ageOlivier Le Meur
 
Saccadic model of eye movements for free-viewing condition
Saccadic model of eye movements for free-viewing conditionSaccadic model of eye movements for free-viewing condition
Saccadic model of eye movements for free-viewing conditionOlivier Le Meur
 
Examplar-based inpainting
Examplar-based inpaintingExamplar-based inpainting
Examplar-based inpaintingOlivier Le Meur
 
Inter-observers congruency and memorability
Inter-observers congruency and memorabilityInter-observers congruency and memorability
Inter-observers congruency and memorabilityOlivier Le Meur
 
Visual attention: models and performance
Visual attention: models and performanceVisual attention: models and performance
Visual attention: models and performanceOlivier Le Meur
 

More from Olivier Le Meur (6)

Guided tour of visual attention
Guided tour of visual attentionGuided tour of visual attention
Guided tour of visual attention
 
Your gaze betrays your age
Your gaze betrays your ageYour gaze betrays your age
Your gaze betrays your age
 
Saccadic model of eye movements for free-viewing condition
Saccadic model of eye movements for free-viewing conditionSaccadic model of eye movements for free-viewing condition
Saccadic model of eye movements for free-viewing condition
 
Examplar-based inpainting
Examplar-based inpaintingExamplar-based inpainting
Examplar-based inpainting
 
Inter-observers congruency and memorability
Inter-observers congruency and memorabilityInter-observers congruency and memorability
Inter-observers congruency and memorability
 
Visual attention: models and performance
Visual attention: models and performanceVisual attention: models and performance
Visual attention: models and performance
 

Recently uploaded

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXssuser89054b
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTbhaskargani46
 
University management System project report..pdf
University management System project report..pdfUniversity management System project report..pdf
University management System project report..pdfKamal Acharya
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756dollysharma2066
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptNANDHAKUMARA10
 
2016EF22_0 solar project report rooftop projects
2016EF22_0 solar project report rooftop projects2016EF22_0 solar project report rooftop projects
2016EF22_0 solar project report rooftop projectssmsksolar
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdfKamal Acharya
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueBhangaleSonal
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaOmar Fathy
 
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Bookingroncy bisnoi
 
A Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityA Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityMorshed Ahmed Rahath
 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VDineshKumar4165
 
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdfONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdfKamal Acharya
 
Unit 2- Effective stress & Permeability.pdf
Unit 2- Effective stress & Permeability.pdfUnit 2- Effective stress & Permeability.pdf
Unit 2- Effective stress & Permeability.pdfRagavanV2
 
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night StandCall Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Standamitlee9823
 
COST-EFFETIVE and Energy Efficient BUILDINGS ptx
COST-EFFETIVE  and Energy Efficient BUILDINGS ptxCOST-EFFETIVE  and Energy Efficient BUILDINGS ptx
COST-EFFETIVE and Energy Efficient BUILDINGS ptxJIT KUMAR GUPTA
 

Recently uploaded (20)

Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar  ≼🔝 Delhi door step de...
Call Now ≽ 9953056974 ≼🔝 Call Girls In New Ashok Nagar ≼🔝 Delhi door step de...
 
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
 
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort ServiceCall Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
Call Girls in Ramesh Nagar Delhi 💯 Call Us 🔝9953056974 🔝 Escort Service
 
Generative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPTGenerative AI or GenAI technology based PPT
Generative AI or GenAI technology based PPT
 
University management System project report..pdf
University management System project report..pdfUniversity management System project report..pdf
University management System project report..pdf
 
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
FULL ENJOY Call Girls In Mahipalpur Delhi Contact Us 8377877756
 
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
(INDIRA) Call Girl Bhosari Call Now 8617697112 Bhosari Escorts 24x7
 
Block diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.pptBlock diagram reduction techniques in control systems.ppt
Block diagram reduction techniques in control systems.ppt
 
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
(INDIRA) Call Girl Meerut Call Now 8617697112 Meerut Escorts 24x7
 
2016EF22_0 solar project report rooftop projects
2016EF22_0 solar project report rooftop projects2016EF22_0 solar project report rooftop projects
2016EF22_0 solar project report rooftop projects
 
Hostel management system project report..pdf
Hostel management system project report..pdfHostel management system project report..pdf
Hostel management system project report..pdf
 
Double Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torqueDouble Revolving field theory-how the rotor develops torque
Double Revolving field theory-how the rotor develops torque
 
Introduction to Serverless with AWS Lambda
Introduction to Serverless with AWS LambdaIntroduction to Serverless with AWS Lambda
Introduction to Serverless with AWS Lambda
 
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance BookingCall Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
Call Girls Wakad Call Me 7737669865 Budget Friendly No Advance Booking
 
A Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna MunicipalityA Study of Urban Area Plan for Pabna Municipality
A Study of Urban Area Plan for Pabna Municipality
 
Thermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - VThermal Engineering-R & A / C - unit - V
Thermal Engineering-R & A / C - unit - V
 
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdfONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
ONLINE FOOD ORDER SYSTEM PROJECT REPORT.pdf
 
Unit 2- Effective stress & Permeability.pdf
Unit 2- Effective stress & Permeability.pdfUnit 2- Effective stress & Permeability.pdf
Unit 2- Effective stress & Permeability.pdf
 
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night StandCall Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
Call Girls In Bangalore ☎ 7737669865 🥵 Book Your One night Stand
 
COST-EFFETIVE and Energy Efficient BUILDINGS ptx
COST-EFFETIVE  and Energy Efficient BUILDINGS ptxCOST-EFFETIVE  and Energy Efficient BUILDINGS ptx
COST-EFFETIVE and Energy Efficient BUILDINGS ptx
 

Color transfer between high-dynamic-range images

  • 1. Color transfer between high- dynamic-range images H. Hristova, R. Cozot, O. Le Meur, K. Bouatouch University of Rennes 1 Rennes, France
  • 2. Outline ● Introduction - Main objective - Contributions ● Extension to the HDR domain of a color transfer method ● Results and evaluation ● Generalization for state-of-the-art color transfer methods ● Conclusion 2
  • 3. Main goal ● Carrying out a color transfer between two HDR images directly in the HDR domain Input Reference 3 ● Solution: apply color transfer methods to stylize an HDR image with regards to a reference image
  • 4. Why do LDR color transfer methods need to be extended to the HDR domain? ● LDR color spaces - well predict the color gamut for luminance levels between zero and the display white point - uncertain applicability to HDR images ● Color trend above the perfect diffuse white 4
  • 5. Why do LDR color transfer methods need to be extended to the HDR domain? ● Assumption: a unique multivariate Gaussian distribution ● HDR domain: to fit the high range of lightness of HDR images we need to assume mixture of Gaussian distributions 5
  • 6. Why do LDR color transfer methods need to be extended to the HDR domain? ● Lightness - approximated by luminance in the LDR domain ● HDR domain - distinguish between the absolute luminance and the lightness (the L channel of CIE Lab) 6
  • 7. Contributions ● Adaptation of [Hristova et al., 2015] color transfer method to HDR images - HDR color spaces - Modifications of the clustering step and of the image classification ● Cluster-based local chromatic adaptation transform ● Generalization for state-of-the-art color transfer methods 7
  • 8. 8 Extension to HDR images • Linear search for significant peaks in the image hue histogram - Colors-based style images: more than one significant color cluster - Light-based style images: one significant color cluster Input and reference images Color space conversion Image classification Clustering and mapping Color transfer Chromatic adaptation transform Final result [Hristovaetal.,2015] • The number of significant peaks determines the number of clusters - Colors-based style images: hue histogram - Light-based style images: luminance histogram
  • 9. Extension to HDR images Input and reference images Color space conversion Image classification Clustering and mapping Color transfer Chromatic adaptation transform Final result 9 [Hristovaetal.,2015] LDR images CIE Lab L channel of CIE Lab HDR images hdr-CIELab Log- luminance ● Dashed line: cubic function of L channel (CIE Lab) ● Solid line: Michaelis-Menten function by which we replace the cubic function of L channel (CIE Lab) ● hdr-CIELab color space [Fairchild et al., 2004] [Fairchild et al., 2004] ModificationsModified
  • 10. Extension to HDR images 10 LDR images CIE Lab L channel of CIE Lab L-based clustering HDR images hdr-CIELab Log- luminance Log- luminance clustering [Hristovaetal.,2015] Logarithmic transform Input and reference images Color space conversion Image classification Clustering and mapping Color transfer Chromatic adaptation transform Final result ModifiedModifications
  • 11. Extension to HDR images 11 Local CAT Cluster- based local CAT [Hristovaetal.,2015] LDR images CIE Lab L channel of CIE Lab L-based clustering HDR images hdr-CIELab Log- luminance Log- luminance clustering Input and reference images Color space conversion Clustering and mapping Color transfer Chromatic adaptation transform Final result Image classification ModifiedModifications
  • 12. Extension to HDR images 12 Gaussianlow-passfilter (h) (m) (sh) (sh) (m) (h) Local CAT Cluster- based local CAT [Hristovaetal.,2015] LDR images CIE Lab L channel of CIE Lab L-based clustering HDR images hdr-CIELab Log- luminance Log- luminance clustering Input and reference images Color space conversion Clustering and mapping Color transfer Chromatic adaptation transform Final result Image classification ModifiedModifications
  • 13. Extension to HDR images 13 Input ReferenceCluster-based local CAT Local CAT Cluster- based local CAT [Hristovaetal.,2015] LDR images CIE Lab L channel of CIE Lab L-based clustering HDR images hdr-CIELab Log- luminance Log- luminance clustering Input and reference images Color space conversion Clustering and mapping Color transfer Chromatic adaptation transform Final result Image classification ModifiedModifications
  • 14. Extension to HDR images 14 Input ReferenceCluster-based local CAT Local CAT Cluster- based local CAT [Hristovaetal.,2015] LDR images CIE Lab L channel of CIE Lab L-based clustering HDR images hdr-CIELab Log- luminance Log- luminance clustering Input and reference images Color space conversion Clustering and mapping Color transfer Chromatic adaptation transform Final result Image classification ModifiedModifications
  • 15. Objective evaluation of the results 15 ● 10 image pairs ● Two tone-mapping operators: [Durand et al., 2002] and [Reinhard et al., 2002] ● SSIM and Bhattacharya coefficient
  • 16. Results 16 Input Reference Color transfer with CAT Color transfer without CAT Color transfer with CAT Color transfer without CAT [Hristova et al., 2015] HDR extension
  • 17. Generalization and results 17 [Reinhard et al., 2001] - global method [Tai et al., 2005] - clustering (local transformations) Input Reference
  • 18. Generalization and results 18 [Pitié et al., 2007] - CIE Lab [Pitié et al., 2007] - hdr-CIELab Input Reference
  • 19. Generalization and results 19 [Bonneel et al., 2013] - luminance clustering [Bonneel et al., 2013] - log-luminance clustering Input Reference
  • 20. Conclusion ● Extension of a novel local color transfer method [Hristova et al., 2015] - Modifications to CIE Lab -> hdr-CIELab - Luminance/Lightness -> Log-luminance ● Generalization to state-of-the-art methods ● Future work - Need for a more precise color mapping/color transformation between two HDR images - Need for better HDR color spaces 20
  • 21. Thank you for your attention! 21

Editor's Notes

  1. Outline of the presentation: Introduction to the main objective and the main contributions of the paper Presenting an extension to the HDR domain of a novel local color transfer method (Followed by) results and evaluation Generalizing the developed extension to other state-of-the-art color transfer methods + results
  2. Our main goal is to carry out a color transfer between two HDR images directly in the HDR domain. -Why? -Because tone mapping operators compress the information contained in HDR images and therefore the plausibility of the tone-mapped images cannot be guaranteed. Solution to the problem: to apply color transfer methods between two HDR images. Color transfer methods build a transformation mapping between two LDR images and aim at “borrowing” the look (mood) of a reference image. Can color transfer methods be directly applied to the HDR domain? -> next slide
  3. The color transfer methods need to be extended to HDR images. -Why? -First: Traditional LDR color spaces, such as lab and CIE Lab, which are widely used in color transfer algorithms, predict well the color gamut for luminance levels below the display white point. However, when it comes to luminance levels way beyond the display white point, like those of HDR images, the applicability of these color spaces becomes uncertain. For example, when the lightness reaches high levels (above 255 in the figure), the imagery primaries (x, y) ranging from 0 to 1 cannot depict and predict the color gamut for these high levels of luminance. To predict the color trend above the perfect diffuse white, we need to perform some modifications to traditional color spaces.
  4. Second: most color transfer methods build their transformations based on the assumption that a multivariate Gaussian distribution is sufficient to fit the luminance and chroma variations. However, this is hardly true in the LDR domain. The latter hold for the HDR domain. The high luminance range of an HDR image cannot be fitted only by one Gaussian distribution. Therefore, we need to assume a mixture of Gaussian distributions.
  5. Finally: Usually, the luminance is approximated by the lightness (the L channel from CIE Lab color space) in the LDR domain. Nevertheless, there is a distinguish between the luminance and the lightness in the HDR domain. On one hand, the luminance of HDR images is presented by the Y channel in XYZ color space and it is called absolute luminance. On the other hand, lightness is represented by the L channel of CIE Lab color space. Both of them are not equal in the context of HDR imagery. Therefore, color transfer methods which use these two notions in their algorithms need to be properly adapted to the HDR domain.
  6. Our main contributions are: the development of extension to HDR images of a new local color transfer method [Hristova et al., 2015], consisting of: - modifications of traditional color spaces (their replacement with HDR-based ones) - modifications of the clustering and classification algorithms of the local method: distinguish between lightness and luminance The development of novel cluster-based local chromatic adaptation transform for adapting the colors of the input HDr image to a reference white point.
  7. The local method by [Hristova et al., 2015] is a new method for color and style transfer consisting of several steps. These steps (either independently or as a combination) can be used in other color transfer methods. Therefore, we will first present the extension of [Hristova et al., 2015] and then, we will generalize that extension to other color transfer methods. [Hristova et al., 2015] starts with a conversion to CIE Lab color space. Once both images are converted to that color space, they are classified according to their main features (either colors or light) as shown on the figure. Two types of images are considered: colors-based and light-based ones. The classification boils down to finding significant peaks in the hue histogram of both images. During the classification step, we determine also the number of clusters which will be passed to the clustering process. On one hand, the number of color clusters is determined by the hue histogram. On the other hand, the number of luminance clusters is defined by the number of significant peaks in the luminance histogram of the image. After the classification, both images are clustered to Gaussian clusters and then, the input/reference clusters are mapped. Four policies are developed to map the clusters. They take into account both the luminance histogram and the luminance-hue distributions of images. A color transfer is carried out between each pair of corresponding clusters. An optimal transformation is used (Monge-Kantorovich optimization problem). Finally, a chromatic adaptation transform is applied to adapt the colors of the output image to the reference illuminant.
  8. The following extension of [Hristova et al., 2015] is proposed. The extension is applied to HDR images instead to LDR images. One of the issues of the direct applying of the method to HDR images has been already discussed: the uncertainty in predicting the color gamut for high luminance values. Therefore, instead of using the LDR CIE Lab color space, we perform modifications to its channels to derive hdr-CIELab color space [Fairchild et al., 2004]. The cubic function of the L channel (the dashed line) is replaced by Michaelis-Menten function (the solid line) for better approximation of the lightness in HDR images. Moreover, this modification influences the predictability of the color trend for HDR images. Furthermore, as the image classification depends strongly on the number of significant peaks in the image hue histogram, if we ensure a better predictability of the image hue, then we ensure the successful classification of images and the successful estimation of the number of color clusters (as the number of significant peaks corresponds to the number of color clusters). How about the number of light/luminance clusters? -> we adopt log-luminance histogram in the place of the L channel histogram -> explanation on the next slide.
  9. To find the optimal number of clusters for an image during the classification step, we need to find number of significant peaks in the luminance histogram of the image. Moreover, light-based images are clustered according to the luminance histogram. Finally, three of the mapping policies are based on luminance histogram information. In the LDR domain we can use the L channel of CIE Lab color space in the place of the luminance. However, in the HDR domain we use the absolute luminance and more precisely, the logarithmic transform of the absolute luminance (which we call log-luminance). The log-luminance is a good approximation of the brightness. Furthermore, the logarithmic transform preserves the regions of minima and maxima of a histogram and therefore we are able to recover the shadows, midtones and highlights from the log-luminance histogram. To this end, log-luminance histogram is used in the classification and clustering steps of the algorithm in the place of the lightness channel of hdr-CIELab color space. Once we carry out the image clustering and we map the reference to the input clusters, we apply an optimal transformation between each pair of corresponding clusters. We call the obtained image an output. The last step adapts the colors of the output image to the reference illuminant.
  10. In the LDR domain, CAT algorithm adapts the input image to a global reference point. The novel cluster-based local CAT algorithm, however, first clusters both images into regions (highlights, midtones, shadows). We search for significant peaks in the log-luminance histogram and then, we define the region limits as the minima between two significant peaks.
  11. Like in the LDR domain, we apply a low-pass Gaussian filter to the input image to compute locally (like a “white” image) the input illuminant. Then, we cluster the input image into regions. Each region corresponds to part of the “white” image which we take into account to carry out the transform region-wisely. As far as the reference image is concerned, we estimate a representative white point for each reference region. Then, this white point becomes a global white point for the region. The reference white points for each region are shown in blue dashed lines.
  12. Here is a video example of how the result changes over the iterations. Around the 10th iteration, the input image is overall adapted to the reference illuminant. However, there is a slight cast of the greenish input reference on the result. It is removed at the end of the iteration.
  13. Another example of illuminant adaptation (video).
  14. We used 10 image pairs to obtain 10 HDR image results for the original [Hristova et al., 2015] method and its extension. We evaluated the results in the LDR domain after tone-mapping them. SSIM and Bhattacharya coefficient are two complementary metrics (used also by [Hristova et al., 2015]) measuring respectively the degree of artifacts and how successful the color transfer is. The HDR extension obtains the highest scores for both metrics and for both tone-mapping operators.
  15. The results support the objective evaluation. The first column represents the input and reference images. The second column shows the results obtained with the method [Hristova et al., 2015] directly applied to HDR images (with and without local CAT). The third column shows the results obtained with the HDR extension of the method (with and without local CAT). If we apply [Hristova et al., 2015] directly to HDR images, then the method it opt for artifacts. Moreover, the details are lost. On the other hand, the HDR extension manages to preserve the details and properly transfers the reference colors to the output image.
  16. Now we show several results for state-of-the-art methods. The first row consists of the input and reference images. The first image on the second row is the result of the direct applying of the state-of-the-art method to HDR images and the second image on the second row presents the result of the same state-of-the-art method, modified in order to enhance the color transfer between two HDR images. [Reinhard et al., 2002] is a global method which assumes that the light and colors of the image can be fitted by a unique Gaussian distribution. As discussed, one Gaussian distribution is not sufficient to fit the high luminance range of HDR images. Therefore, to enhance the effect of the color transfer, we recommend to carry out Reinhard’s transformation locally on pairs of clusters. To this end, we recommend [Tai et al., 2005] to be used in the place of [Reinhard et al., 2005] in the HDR domain.
  17. [Pitié et al., 2007] is a global method which applies optimal mapping between the input and reference images. In the LDR domain, the methods adopts CIE Lab color space. As we can see from the first result on the figure, CIE Lab is less effective from this type of transformation between HDR images than hdr-CIELab. The latter color space predicts better the color palette of the reference image and manages to transfer it better to the final result.
  18. As a local method, [Bonneel et al., 2013] performs clustering to both input and reference images. The clustering is luminance-based. To extend the method to HDR images, we replace the luminance-based clustering with log-luminance-based clustering (as shown for [Hristova et al., 2015] earlier in the presentation). Moreover, instead of carrying out the transformation in CIE Lab color space, we adopt hdr-CIELab color space in the HDR domain. That way, the proposed extension lessens the visual artifacts in the final result (in comparison to the direct application of [Bonneel et al., 2013]).