SlideShare a Scribd company logo
18 January 2018 IEEE -Tap Energy 2017 1
Single Image Super Resolution using Fuzzy
Deep Convolutional Networks
Presented By
Greeshma M.S
School of Computer Sciences Mahatma Gandhi University,
Kottayam, Kerala, India
Authors
Greeshma M S, Dr. Bindu V R
Paper Id : 145
December 22, 2017
International (biennial) Conference on Technological Advancements in Power & Energy– TAP Energy 2017, Amrita University
18 January 2018 IEEE -Tap Energy 2017 2
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Want to Make Image Big &
Sharp”
Image Super –Resolution (SR)
Given a low resolution image , how do we make it larger without losing any details?
Low-Resolution
High-Resolution
SUPER
RESOLUTION
Applications
• Satellite Imaging
• Microscopy
• HD video generation from low-
resolution source
• Medical Imaging
• Object/ Text recognition from
Surveillance video
• ……………….
18 January 2018 IEEE -Tap Energy 2017 3
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
The “Playground”
Single Image Super Resolution (SISR)
Single Image
Super Resolution
Algorithm
Scale Factor
Image
Model
Manifold Learning Dictionary Learning Deep Learning
Very Challenging !
Are the generated details real?
18 January 2018 IEEE -Tap Energy 2017 4
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
The “Playground”
Problem Domain
Well designed efficient algorithm for increasing image resolution, maintaining better visual quality
and preserving information using Fuzzy Deep Convolutional Network.
Old but Still Hot : Several decades ago [Huang et al] near recent
Many real world applications
Convolutional Neural N1etwork: recent developments
Motivation
18 January 2018 IEEE -Tap Energy 2017 5
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Related works”
Conventional SISR Algorithms
1. Manifold learning
2. Dictionary learning
3. Deep learning
• Effectiveness
• Model issues
• Slow Performance
Remaining Challenges
18 January 2018 IEEE -Tap Energy 2017 6
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep
Convolutional Networks
Experimental Results
Conclusion
“Proposed Network”
Single Image Super Resolution using Fuzzy Deep Convolutional Networks
Utilize Contextual information spread over very large image region
We employ deep CNN model
To keep pixel-wise info only
Directly learns an end-to-end feature mapping between the low to high-
resolution image.
Key contribution:: incorporation of a fuzzy rule layer with the CNN
structure
Fuzzy rule layer added to CNN
network offers:
• Noise reduction property
• Task-driven feature learning
(preserve spatial coherence).
• Rule-driven Selective patch
processing
18 January 2018 IEEE -Tap Energy 2017 7
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep
Convolutional Networks
Experimental Results
Conclusion
“Proposed Network”
Single Image Super Resolution using Fuzzy Deep Convolutional Networks
Our approach for fast convergence and scales
Fuzzy CNN performs hierarchical feature learning between LR image
and final HR image
 Fuzzy deep learning SISR approach comprises of a two-level
reconstruction schema,
(i) Model Construction
(ii) Image Reconstruction
Advantages
• Fast Convergence
• Multi-scale- Train a single
convolutional network to learn and
handle multi-scale factors
Different scale HELPS each other!
18 January 2018 IEEE -Tap Energy 2017 8
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep
Convolutional Networks
Experimental Results
Conclusion
“Proposed Network”
Single Image Super Resolution using Fuzzy Deep Convolutional Networks
1. Model Construction
 Cuda-convnet package (caffe package and prototxt) and the caffe
models provide end-to-end machine learning systems
 Generate LR and HR patches from training images
 Sample network model
(fsub−f1−f2−f3+3)2 =(9-1-5) = (9+5−1)2
=169 pixels
 In training , Goal is to minimize
1
𝑛 𝑖=1
𝑛
𝐹 𝑋𝑖; 𝛩 − 𝑌𝑖 2
 Enable High learning rate
SPEC
1. 93 training images
2. Size of sub images=33
3. f1=9, f2=1,f3=5
4. n1=64, n2=32
5. 3 Convolutional layers
6. 3×3 filters, 64 channels
18 January 2018 IEEE -Tap Energy 2017 9
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep
Convolutional Networks
Experimental Results
Conclusion
“Proposed Network”
Single Image Super Resolution using Fuzzy Deep Convolutional Networks
2. Image Reconstruction
 Image is reconstructed using the network model to enhance the low-
resolution image.
 The key phases to reconstruct the HR image using fuzzy CNN
Feature extraction and representation
Non-linearity mapping of layers
Accumulation of ultimate layer and fuzzy rule layer to reconstruct
the HR image
Fuzzy rule layer
Feature maps extracted by the
conv1 layer may contain hand-
crafted features or ambiguities, and
hence the fuzzy rule layer is
inserted to stifle anomalies or
contaminations and to enhance
informative features.
18 January 2018 IEEE -Tap Energy 2017 10
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep
Convolutional Networks
Experimental Results
Conclusion
“Proposed Network”
Processing Pipeline
Mathematical repesentation
If x is ADJACENT to HR=
𝑔𝑐
𝑥
≈
𝑖𝑓 𝑥 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑐
𝑥
𝐴𝑁𝐷 𝑖𝑓 𝑥2 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑐
𝑥
…..…
𝐴𝑁𝐷 𝑖𝑓 𝑥𝑝 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑛
𝑥
18 January 2018 IEEE -Tap Energy 2017 11
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
Bicubic
Interpolation
Our
approach
VS.
Sharp
Low
visual
Quality
Even
Sharper
Richer
texture
Visually
pleasing
Detailed
image
Sparse
coding
18 January 2018 IEEE -Tap Energy 2017 12
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
18 January 2018 IEEE -Tap Energy 2017 13
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
 Cuda-convnet package used for train a model
 Qualitatively and quantitatively analyzed by;
Peak Signal Noise Ratio (PSNR)
Structural Similarity Index( SSIM)
Feature Similarity Index (FSIM )
 Performance of the auto-learning algorithm compared with bicubic interpolation, Sparse coding, and SRCNN
One model, Multiple-scales
18 January 2018 IEEE -Tap Energy 2017 14
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
Bicubic interpolation Sparse coding
SRCNN
3
Fuzzy Deep
Learning
Quantitative Analysis
18 January 2018 IEEE -Tap Energy 2017
15
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
Bicubic interpolation Sparse coding
SRCNN
3
Fuzzy Deep
Learning
Quantitative Analysis
Magnifica
tion
Factor
Algorithms
Measures Bicubic Sparse
Coding
SRCNN Fuzzy
Deep
Learning
3
PSNR(dB)
SSIM
FSIM
27.36
0.744
0.818
27.77
0.770
0.834
28.12
0.792
0.839
28.47
0.810
0.858
18 January 2018 IEEE -Tap Energy 2017
16
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion
“Experiment Results”
Bicubic interpolation Sparse coding
SRCNN
3
Fuzzy Deep
Learning
Quantitative Analysis
Magnifica
tion
Factor
Algorithms
Measures Bicubic Sparse
Coding
SRCNN Fuzzy
Deep
Learning
3
PSNR(dB)
SSIM
FSIM
24.78
0.765
0.854
25.68
0.805
0.885
26.49
0.861
0.928
28.01
0.889
0.943
18 January 2018 IEEE -Tap Energy 2017
17
Introduction to super resolution problem
conventional Approaches
Single Image Super Resolution using Fuzzy Deep Convolutional
Networks
Experimental Results
Conclusion“Findings”
CONCLUSION
•Avoiding additional overhead during learning such as weight sharing, pooling etc.
• Highlighting feature: fuzzy rule layer accumulates with CNN and offers selective patch processing and noise
reduction
• Ability of the proposed method to preserve the structural information in the final HR image with better visual
quality
•Interested in exploring the applicability of proposed method to video super resolution and audio super
resolution
18 January 2018 IEEE -Tap Energy 2017
18
Thank You
International (biennial) Conference on Technological Advancements in Power & Energy– TAP Energy 2017, Amrita University

More Related Content

What's hot

Literature Review on Single Image Super Resolution
Literature Review on Single Image Super ResolutionLiterature Review on Single Image Super Resolution
Literature Review on Single Image Super Resolution
ijtsrd
 
Image super resolution
Image super resolutionImage super resolution
Image super resolution
Akshay Hazare
 
Deep-Learning Based Stereo Super-Resolution
Deep-Learning Based Stereo Super-ResolutionDeep-Learning Based Stereo Super-Resolution
Deep-Learning Based Stereo Super-Resolution
NAVER Engineering
 
Super resolution in deep learning era - Jaejun Yoo
Super resolution in deep learning era - Jaejun YooSuper resolution in deep learning era - Jaejun Yoo
Super resolution in deep learning era - Jaejun Yoo
JaeJun Yoo
 
Image super resolution based on
Image super resolution based onImage super resolution based on
Image super resolution based on
jpstudcorner
 
Deep learning for image super resolution
Deep learning for image super resolutionDeep learning for image super resolution
Deep learning for image super resolution
Prudhvi Raj
 
2019 cvpr paper_overview
2019 cvpr paper_overview2019 cvpr paper_overview
2019 cvpr paper_overview
LEE HOSEONG
 
(Research Note) Delving deeper into convolutional neural networks for camera ...
(Research Note) Delving deeper into convolutional neural networks for camera ...(Research Note) Delving deeper into convolutional neural networks for camera ...
(Research Note) Delving deeper into convolutional neural networks for camera ...
Jacky Liu
 
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen..."The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
LEE HOSEONG
 
Super Resolution
Super ResolutionSuper Resolution
Super Resolution
alokahuti
 
[CVPR2020] Simple but effective image enhancement techniques
[CVPR2020] Simple but effective image enhancement techniques[CVPR2020] Simple but effective image enhancement techniques
[CVPR2020] Simple but effective image enhancement techniques
JaeJun Yoo
 
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
Universitat Politècnica de Catalunya
 
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
Wanjin Yu
 
AlexNet
AlexNetAlexNet
AlexNet
Bertil Hatt
 
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
Fellowship at Vodafone FutureLab
 
Modeling perceptual similarity and shift invariance in deep networks
Modeling perceptual similarity and shift invariance in deep networksModeling perceptual similarity and shift invariance in deep networks
Modeling perceptual similarity and shift invariance in deep networks
NAVER Engineering
 
Lecture 29 Convolutional Neural Networks - Computer Vision Spring2015
Lecture 29 Convolutional Neural Networks -  Computer Vision Spring2015Lecture 29 Convolutional Neural Networks -  Computer Vision Spring2015
Lecture 29 Convolutional Neural Networks - Computer Vision Spring2015
Jia-Bin Huang
 
[PR12] Generative Models as Distributions of Functions
[PR12] Generative Models as Distributions of Functions[PR12] Generative Models as Distributions of Functions
[PR12] Generative Models as Distributions of Functions
JaeJun Yoo
 
AI&BigData Lab. Артем Чернодуб "Распознавание изображений методом Lazy Deep ...
AI&BigData Lab. Артем Чернодуб  "Распознавание изображений методом Lazy Deep ...AI&BigData Lab. Артем Чернодуб  "Распознавание изображений методом Lazy Deep ...
AI&BigData Lab. Артем Чернодуб "Распознавание изображений методом Lazy Deep ...
GeeksLab Odessa
 
Super resolution-review
Super resolution-reviewSuper resolution-review
Super resolution-review
Woojin Jeong
 

What's hot (20)

Literature Review on Single Image Super Resolution
Literature Review on Single Image Super ResolutionLiterature Review on Single Image Super Resolution
Literature Review on Single Image Super Resolution
 
Image super resolution
Image super resolutionImage super resolution
Image super resolution
 
Deep-Learning Based Stereo Super-Resolution
Deep-Learning Based Stereo Super-ResolutionDeep-Learning Based Stereo Super-Resolution
Deep-Learning Based Stereo Super-Resolution
 
Super resolution in deep learning era - Jaejun Yoo
Super resolution in deep learning era - Jaejun YooSuper resolution in deep learning era - Jaejun Yoo
Super resolution in deep learning era - Jaejun Yoo
 
Image super resolution based on
Image super resolution based onImage super resolution based on
Image super resolution based on
 
Deep learning for image super resolution
Deep learning for image super resolutionDeep learning for image super resolution
Deep learning for image super resolution
 
2019 cvpr paper_overview
2019 cvpr paper_overview2019 cvpr paper_overview
2019 cvpr paper_overview
 
(Research Note) Delving deeper into convolutional neural networks for camera ...
(Research Note) Delving deeper into convolutional neural networks for camera ...(Research Note) Delving deeper into convolutional neural networks for camera ...
(Research Note) Delving deeper into convolutional neural networks for camera ...
 
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen..."The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
"The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Gen...
 
Super Resolution
Super ResolutionSuper Resolution
Super Resolution
 
[CVPR2020] Simple but effective image enhancement techniques
[CVPR2020] Simple but effective image enhancement techniques[CVPR2020] Simple but effective image enhancement techniques
[CVPR2020] Simple but effective image enhancement techniques
 
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
Visual Saliency Prediction with Deep Learning - Kevin McGuinness - UPC Barcel...
 
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
Intelligent Image Enhancement and Restoration - From Prior Driven Model to Ad...
 
AlexNet
AlexNetAlexNet
AlexNet
 
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
AlexNet(ImageNet Classification with Deep Convolutional Neural Networks)
 
Modeling perceptual similarity and shift invariance in deep networks
Modeling perceptual similarity and shift invariance in deep networksModeling perceptual similarity and shift invariance in deep networks
Modeling perceptual similarity and shift invariance in deep networks
 
Lecture 29 Convolutional Neural Networks - Computer Vision Spring2015
Lecture 29 Convolutional Neural Networks -  Computer Vision Spring2015Lecture 29 Convolutional Neural Networks -  Computer Vision Spring2015
Lecture 29 Convolutional Neural Networks - Computer Vision Spring2015
 
[PR12] Generative Models as Distributions of Functions
[PR12] Generative Models as Distributions of Functions[PR12] Generative Models as Distributions of Functions
[PR12] Generative Models as Distributions of Functions
 
AI&BigData Lab. Артем Чернодуб "Распознавание изображений методом Lazy Deep ...
AI&BigData Lab. Артем Чернодуб  "Распознавание изображений методом Lazy Deep ...AI&BigData Lab. Артем Чернодуб  "Распознавание изображений методом Lazy Deep ...
AI&BigData Lab. Артем Чернодуб "Распознавание изображений методом Lazy Deep ...
 
Super resolution-review
Super resolution-reviewSuper resolution-review
Super resolution-review
 

Similar to Single Image Super Resolution using Fuzzy Deep Convolutional Networks

Pixel Recursive Super Resolution. Google Brain
 Pixel Recursive Super Resolution.  Google Brain Pixel Recursive Super Resolution.  Google Brain
Pixel Recursive Super Resolution. Google Brain
eraser Juan José Calderón
 
Full resolution image compression with recurrent neural networks
Full resolution image compression with recurrent neural networksFull resolution image compression with recurrent neural networks
Full resolution image compression with recurrent neural networks
Ashis Chanda
 
Full resolution image compression with recurrent neural networks
Full resolution image compression with  recurrent neural networksFull resolution image compression with  recurrent neural networks
Full resolution image compression with recurrent neural networks
Ashis Kumar Chanda
 
IRJET- Exploring Image Super Resolution Techniques
IRJET- Exploring Image Super Resolution TechniquesIRJET- Exploring Image Super Resolution Techniques
IRJET- Exploring Image Super Resolution Techniques
IRJET Journal
 
Deep learning for image super resolution
Deep learning for image super resolutionDeep learning for image super resolution
Deep learning for image super resolution
Prudhvi Raj
 
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
sipij
 
Decomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesisDecomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesis
Naeem Shehzad
 
Pixel Recurrent Neural Networks
Pixel Recurrent Neural NetworksPixel Recurrent Neural Networks
Pixel Recurrent Neural Networks
neouyghur
 
Image super resolution using Generative Adversarial Network.
Image super resolution using Generative Adversarial Network.Image super resolution using Generative Adversarial Network.
Image super resolution using Generative Adversarial Network.
IRJET Journal
 
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
sipij
 
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
sipij
 
Single Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
Single Image Super-Resolution Using Analytical Solution for L2-L2 AlgorithmSingle Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
Single Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
ijtsrd
 
Inpainting related works (part 2)
Inpainting related works (part 2)Inpainting related works (part 2)
Inpainting related works (part 2)
Seowoo Han
 
Hyper-parameter optimization of convolutional neural network based on particl...
Hyper-parameter optimization of convolutional neural network based on particl...Hyper-parameter optimization of convolutional neural network based on particl...
Hyper-parameter optimization of convolutional neural network based on particl...
journalBEEI
 
Mnist report
Mnist reportMnist report
Mnist report
RaghunandanJairam
 
Mnist report ppt
Mnist report pptMnist report ppt
Mnist report ppt
RaghunandanJairam
 
Super-Resolution of Multispectral Images
Super-Resolution of Multispectral ImagesSuper-Resolution of Multispectral Images
Super-Resolution of Multispectral Images
ijsrd.com
 
Seminarpaper
SeminarpaperSeminarpaper
Seminarpaper
PrashantChaudhari75
 
Image De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural NetworkImage De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural Network
aciijournal
 
Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation
Sneha Ravikumar
 

Similar to Single Image Super Resolution using Fuzzy Deep Convolutional Networks (20)

Pixel Recursive Super Resolution. Google Brain
 Pixel Recursive Super Resolution.  Google Brain Pixel Recursive Super Resolution.  Google Brain
Pixel Recursive Super Resolution. Google Brain
 
Full resolution image compression with recurrent neural networks
Full resolution image compression with recurrent neural networksFull resolution image compression with recurrent neural networks
Full resolution image compression with recurrent neural networks
 
Full resolution image compression with recurrent neural networks
Full resolution image compression with  recurrent neural networksFull resolution image compression with  recurrent neural networks
Full resolution image compression with recurrent neural networks
 
IRJET- Exploring Image Super Resolution Techniques
IRJET- Exploring Image Super Resolution TechniquesIRJET- Exploring Image Super Resolution Techniques
IRJET- Exploring Image Super Resolution Techniques
 
Deep learning for image super resolution
Deep learning for image super resolutionDeep learning for image super resolution
Deep learning for image super resolution
 
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
ADVANCED SINGLE IMAGE RESOLUTION UPSURGING USING A GENERATIVE ADVERSARIAL NET...
 
Decomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesisDecomposing image generation into layout priction and conditional synthesis
Decomposing image generation into layout priction and conditional synthesis
 
Pixel Recurrent Neural Networks
Pixel Recurrent Neural NetworksPixel Recurrent Neural Networks
Pixel Recurrent Neural Networks
 
Image super resolution using Generative Adversarial Network.
Image super resolution using Generative Adversarial Network.Image super resolution using Generative Adversarial Network.
Image super resolution using Generative Adversarial Network.
 
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
 
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENTAPPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
APPLICATION OF CONVOLUTIONAL NEURAL NETWORK IN LAWN MEASUREMENT
 
Single Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
Single Image Super-Resolution Using Analytical Solution for L2-L2 AlgorithmSingle Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
Single Image Super-Resolution Using Analytical Solution for L2-L2 Algorithm
 
Inpainting related works (part 2)
Inpainting related works (part 2)Inpainting related works (part 2)
Inpainting related works (part 2)
 
Hyper-parameter optimization of convolutional neural network based on particl...
Hyper-parameter optimization of convolutional neural network based on particl...Hyper-parameter optimization of convolutional neural network based on particl...
Hyper-parameter optimization of convolutional neural network based on particl...
 
Mnist report
Mnist reportMnist report
Mnist report
 
Mnist report ppt
Mnist report pptMnist report ppt
Mnist report ppt
 
Super-Resolution of Multispectral Images
Super-Resolution of Multispectral ImagesSuper-Resolution of Multispectral Images
Super-Resolution of Multispectral Images
 
Seminarpaper
SeminarpaperSeminarpaper
Seminarpaper
 
Image De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural NetworkImage De-Noising Using Deep Neural Network
Image De-Noising Using Deep Neural Network
 
Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation Ultrasound Nerve Segmentation
Ultrasound Nerve Segmentation
 

Recently uploaded

MARY JANE WILSON, A “BOA MÃE” .
MARY JANE WILSON, A “BOA MÃE”           .MARY JANE WILSON, A “BOA MÃE”           .
MARY JANE WILSON, A “BOA MÃE” .
Colégio Santa Teresinha
 
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
Nguyen Thanh Tu Collection
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
Jean Carlos Nunes Paixão
 
How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17
Celine George
 
South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)
Academy of Science of South Africa
 
The Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collectionThe Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collection
Israel Genealogy Research Association
 
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
Dr. Vinod Kumar Kanvaria
 
Main Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docxMain Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docx
adhitya5119
 
PIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf IslamabadPIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf Islamabad
AyyanKhan40
 
The basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptxThe basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptx
heathfieldcps1
 
Natural birth techniques - Mrs.Akanksha Trivedi Rama University
Natural birth techniques - Mrs.Akanksha Trivedi Rama UniversityNatural birth techniques - Mrs.Akanksha Trivedi Rama University
Natural birth techniques - Mrs.Akanksha Trivedi Rama University
Akanksha trivedi rama nursing college kanpur.
 
Liberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdfLiberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdf
WaniBasim
 
Film vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movieFilm vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movie
Nicholas Montgomery
 
Assessment and Planning in Educational technology.pptx
Assessment and Planning in Educational technology.pptxAssessment and Planning in Educational technology.pptx
Assessment and Planning in Educational technology.pptx
Kavitha Krishnan
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
simonomuemu
 
How to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold MethodHow to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold Method
Celine George
 
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdfANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
Priyankaranawat4
 
S1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptxS1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptx
tarandeep35
 
World environment day ppt For 5 June 2024
World environment day ppt For 5 June 2024World environment day ppt For 5 June 2024
World environment day ppt For 5 June 2024
ak6969907
 
CACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdfCACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdf
camakaiclarkmusic
 

Recently uploaded (20)

MARY JANE WILSON, A “BOA MÃE” .
MARY JANE WILSON, A “BOA MÃE”           .MARY JANE WILSON, A “BOA MÃE”           .
MARY JANE WILSON, A “BOA MÃE” .
 
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
BÀI TẬP BỔ TRỢ TIẾNG ANH 8 CẢ NĂM - GLOBAL SUCCESS - NĂM HỌC 2023-2024 (CÓ FI...
 
Lapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdfLapbook sobre os Regimes Totalitários.pdf
Lapbook sobre os Regimes Totalitários.pdf
 
How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17How to Fix the Import Error in the Odoo 17
How to Fix the Import Error in the Odoo 17
 
South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)South African Journal of Science: Writing with integrity workshop (2024)
South African Journal of Science: Writing with integrity workshop (2024)
 
The Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collectionThe Diamonds of 2023-2024 in the IGRA collection
The Diamonds of 2023-2024 in the IGRA collection
 
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
Exploiting Artificial Intelligence for Empowering Researchers and Faculty, In...
 
Main Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docxMain Java[All of the Base Concepts}.docx
Main Java[All of the Base Concepts}.docx
 
PIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf IslamabadPIMS Job Advertisement 2024.pdf Islamabad
PIMS Job Advertisement 2024.pdf Islamabad
 
The basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptxThe basics of sentences session 6pptx.pptx
The basics of sentences session 6pptx.pptx
 
Natural birth techniques - Mrs.Akanksha Trivedi Rama University
Natural birth techniques - Mrs.Akanksha Trivedi Rama UniversityNatural birth techniques - Mrs.Akanksha Trivedi Rama University
Natural birth techniques - Mrs.Akanksha Trivedi Rama University
 
Liberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdfLiberal Approach to the Study of Indian Politics.pdf
Liberal Approach to the Study of Indian Politics.pdf
 
Film vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movieFilm vocab for eal 3 students: Australia the movie
Film vocab for eal 3 students: Australia the movie
 
Assessment and Planning in Educational technology.pptx
Assessment and Planning in Educational technology.pptxAssessment and Planning in Educational technology.pptx
Assessment and Planning in Educational technology.pptx
 
Smart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICTSmart-Money for SMC traders good time and ICT
Smart-Money for SMC traders good time and ICT
 
How to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold MethodHow to Build a Module in Odoo 17 Using the Scaffold Method
How to Build a Module in Odoo 17 Using the Scaffold Method
 
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdfANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
ANATOMY AND BIOMECHANICS OF HIP JOINT.pdf
 
S1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptxS1-Introduction-Biopesticides in ICM.pptx
S1-Introduction-Biopesticides in ICM.pptx
 
World environment day ppt For 5 June 2024
World environment day ppt For 5 June 2024World environment day ppt For 5 June 2024
World environment day ppt For 5 June 2024
 
CACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdfCACJapan - GROUP Presentation 1- Wk 4.pdf
CACJapan - GROUP Presentation 1- Wk 4.pdf
 

Single Image Super Resolution using Fuzzy Deep Convolutional Networks

  • 1. 18 January 2018 IEEE -Tap Energy 2017 1 Single Image Super Resolution using Fuzzy Deep Convolutional Networks Presented By Greeshma M.S School of Computer Sciences Mahatma Gandhi University, Kottayam, Kerala, India Authors Greeshma M S, Dr. Bindu V R Paper Id : 145 December 22, 2017 International (biennial) Conference on Technological Advancements in Power & Energy– TAP Energy 2017, Amrita University
  • 2. 18 January 2018 IEEE -Tap Energy 2017 2 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Want to Make Image Big & Sharp” Image Super –Resolution (SR) Given a low resolution image , how do we make it larger without losing any details? Low-Resolution High-Resolution SUPER RESOLUTION Applications • Satellite Imaging • Microscopy • HD video generation from low- resolution source • Medical Imaging • Object/ Text recognition from Surveillance video • ……………….
  • 3. 18 January 2018 IEEE -Tap Energy 2017 3 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion The “Playground” Single Image Super Resolution (SISR) Single Image Super Resolution Algorithm Scale Factor Image Model Manifold Learning Dictionary Learning Deep Learning Very Challenging ! Are the generated details real?
  • 4. 18 January 2018 IEEE -Tap Energy 2017 4 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion The “Playground” Problem Domain Well designed efficient algorithm for increasing image resolution, maintaining better visual quality and preserving information using Fuzzy Deep Convolutional Network. Old but Still Hot : Several decades ago [Huang et al] near recent Many real world applications Convolutional Neural N1etwork: recent developments Motivation
  • 5. 18 January 2018 IEEE -Tap Energy 2017 5 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Related works” Conventional SISR Algorithms 1. Manifold learning 2. Dictionary learning 3. Deep learning • Effectiveness • Model issues • Slow Performance Remaining Challenges
  • 6. 18 January 2018 IEEE -Tap Energy 2017 6 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Proposed Network” Single Image Super Resolution using Fuzzy Deep Convolutional Networks Utilize Contextual information spread over very large image region We employ deep CNN model To keep pixel-wise info only Directly learns an end-to-end feature mapping between the low to high- resolution image. Key contribution:: incorporation of a fuzzy rule layer with the CNN structure Fuzzy rule layer added to CNN network offers: • Noise reduction property • Task-driven feature learning (preserve spatial coherence). • Rule-driven Selective patch processing
  • 7. 18 January 2018 IEEE -Tap Energy 2017 7 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Proposed Network” Single Image Super Resolution using Fuzzy Deep Convolutional Networks Our approach for fast convergence and scales Fuzzy CNN performs hierarchical feature learning between LR image and final HR image  Fuzzy deep learning SISR approach comprises of a two-level reconstruction schema, (i) Model Construction (ii) Image Reconstruction Advantages • Fast Convergence • Multi-scale- Train a single convolutional network to learn and handle multi-scale factors Different scale HELPS each other!
  • 8. 18 January 2018 IEEE -Tap Energy 2017 8 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Proposed Network” Single Image Super Resolution using Fuzzy Deep Convolutional Networks 1. Model Construction  Cuda-convnet package (caffe package and prototxt) and the caffe models provide end-to-end machine learning systems  Generate LR and HR patches from training images  Sample network model (fsub−f1−f2−f3+3)2 =(9-1-5) = (9+5−1)2 =169 pixels  In training , Goal is to minimize 1 𝑛 𝑖=1 𝑛 𝐹 𝑋𝑖; 𝛩 − 𝑌𝑖 2  Enable High learning rate SPEC 1. 93 training images 2. Size of sub images=33 3. f1=9, f2=1,f3=5 4. n1=64, n2=32 5. 3 Convolutional layers 6. 3×3 filters, 64 channels
  • 9. 18 January 2018 IEEE -Tap Energy 2017 9 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Proposed Network” Single Image Super Resolution using Fuzzy Deep Convolutional Networks 2. Image Reconstruction  Image is reconstructed using the network model to enhance the low- resolution image.  The key phases to reconstruct the HR image using fuzzy CNN Feature extraction and representation Non-linearity mapping of layers Accumulation of ultimate layer and fuzzy rule layer to reconstruct the HR image Fuzzy rule layer Feature maps extracted by the conv1 layer may contain hand- crafted features or ambiguities, and hence the fuzzy rule layer is inserted to stifle anomalies or contaminations and to enhance informative features.
  • 10. 18 January 2018 IEEE -Tap Energy 2017 10 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Proposed Network” Processing Pipeline Mathematical repesentation If x is ADJACENT to HR= 𝑔𝑐 𝑥 ≈ 𝑖𝑓 𝑥 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑐 𝑥 𝐴𝑁𝐷 𝑖𝑓 𝑥2 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑐 𝑥 …..… 𝐴𝑁𝐷 𝑖𝑓 𝑥𝑝 𝑖𝑠 𝐴𝐷𝐽𝐴𝐶𝐸𝑁𝑇 𝑡𝑜 𝑔 𝑛 𝑥
  • 11. 18 January 2018 IEEE -Tap Energy 2017 11 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results” Bicubic Interpolation Our approach VS. Sharp Low visual Quality Even Sharper Richer texture Visually pleasing Detailed image Sparse coding
  • 12. 18 January 2018 IEEE -Tap Energy 2017 12 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results”
  • 13. 18 January 2018 IEEE -Tap Energy 2017 13 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results”  Cuda-convnet package used for train a model  Qualitatively and quantitatively analyzed by; Peak Signal Noise Ratio (PSNR) Structural Similarity Index( SSIM) Feature Similarity Index (FSIM )  Performance of the auto-learning algorithm compared with bicubic interpolation, Sparse coding, and SRCNN One model, Multiple-scales
  • 14. 18 January 2018 IEEE -Tap Energy 2017 14 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results” Bicubic interpolation Sparse coding SRCNN 3 Fuzzy Deep Learning Quantitative Analysis
  • 15. 18 January 2018 IEEE -Tap Energy 2017 15 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results” Bicubic interpolation Sparse coding SRCNN 3 Fuzzy Deep Learning Quantitative Analysis Magnifica tion Factor Algorithms Measures Bicubic Sparse Coding SRCNN Fuzzy Deep Learning 3 PSNR(dB) SSIM FSIM 27.36 0.744 0.818 27.77 0.770 0.834 28.12 0.792 0.839 28.47 0.810 0.858
  • 16. 18 January 2018 IEEE -Tap Energy 2017 16 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion “Experiment Results” Bicubic interpolation Sparse coding SRCNN 3 Fuzzy Deep Learning Quantitative Analysis Magnifica tion Factor Algorithms Measures Bicubic Sparse Coding SRCNN Fuzzy Deep Learning 3 PSNR(dB) SSIM FSIM 24.78 0.765 0.854 25.68 0.805 0.885 26.49 0.861 0.928 28.01 0.889 0.943
  • 17. 18 January 2018 IEEE -Tap Energy 2017 17 Introduction to super resolution problem conventional Approaches Single Image Super Resolution using Fuzzy Deep Convolutional Networks Experimental Results Conclusion“Findings” CONCLUSION •Avoiding additional overhead during learning such as weight sharing, pooling etc. • Highlighting feature: fuzzy rule layer accumulates with CNN and offers selective patch processing and noise reduction • Ability of the proposed method to preserve the structural information in the final HR image with better visual quality •Interested in exploring the applicability of proposed method to video super resolution and audio super resolution
  • 18. 18 January 2018 IEEE -Tap Energy 2017 18 Thank You International (biennial) Conference on Technological Advancements in Power & Energy– TAP Energy 2017, Amrita University