Image super resolution using Generative Adversarial Network.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 05 | May 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1498
Image super resolution using Generative Adversarial Network.
Vinaya Akhade
1
, Rutuja Naikwade
2
, Harshada Sherkar
3
,Mrs. Asmita R. Kamble
4
1,2,3,4 B.Tech student, Computer Science, Sinhgad Institute of Technology and Science , Pune, India
---------------------------------------------------------------------------***--------------------------------------------------------------------------
Abstract -
Super-resolution (SR) is an image processing technique thataims to increase the resolution of an image by adding sub- pixel
detail. The information used for adding detail can come from sub-pixel shifts provided by sequences of images (frequency
domain), or by a good understanding of the degradation processes, including blurring, that cause the loss of detail.
Convolutional neural networks (CNNs) are especially suited for this type of application due to their ability to empirically
map the underlying connections between an image pixel and those surrounding it. Conversionfrom multiple low resolution
(LR) images to high resolution (HR) image is done by using super-resolution techniques. Anyone can achieve more
information in detail from high- resolution images, which helps further for many satellite image applications. This growing
technology interest in the reconstruction of imagery leads to several methodologies in the field of advanced digital color
image processing.
Recent years have seen growing interest in the problem of super-resolution restoration of video sequences. Whereas in the
traditional single image restoration problem only a single input image is available for processing, the task of reconstructing
super-resolution images from multiple under sampled and degraded images can take advantage of the additional patio-
temporal data available in the image sequence. In particular, camera and scene motion lead to frames in the source video
sequence containing similar, but not identical information. The additional information available in these frames make
possible reconstruction of visually superior frames at higher resolution than that of theoriginal data.
Key Words: Convolutional Neural Network (CNN), Super-Resolution (SR), High-resolution (HR) , Low resolution
(LR)
1.INTRODUCTION
Image Super-Resolution (SR) is an important class of image processing techniques to enhance the resolution ofimages
and videos in computer vision. Recent years have witnessed remarkable progress of image super-resolution using deep
learning techniques. Super resolution is the process of combining a sequence of low-resolution (LR) noisy blurred
images to produce a higher resolution image or sequence. Super- resolution of image is the most widely used and extensive
area of research. The resolution is referred as an important aspect of image. The problem of limited resolution by image
acquisition devices can be solved by super resolution.
Image super-resolution (SR), which refers to the processof recovering high-resolution (HR) images from low resolution
(LR) images, is an important class of image processing techniques in computer vision and image processing. It enjoys a
wide range of real-world applications, such as medical imaging, surveillance and security , amongst others. Other than
improving image perceptual quality, it also helps to improve other computer vision tasks. In general, this problem is
very challenging and inherently ill-posed since there are always multiple HR images corresponding to a single LRimage.
In literature, a variety of classical SR methods have been proposed, including prediction-based methods , edge-based
methods ,statistical methods, patch-based methods and sparse representation methods etc.
The main contributions of this survey are three-fold: 1) We give a comprehensive review of image super resolution
techniques based on deep learning, including problem settings, benchmark datasets, performancemetrics, a family of SR
methods with deep learning, domain-specific SR applications, etc. 2) We provide a systematic overview of recent
advances of deep learning based SR techniques in a hierarchical and structuralmanner, and summarize the advantages

and limitationsof each component for an effective SR solution. 3) We discuss the challenges and open issues, and identify
thenew trends and future directions to provide an insightful guidance for the community.
2. Literature Review
Priyanka and Rishabh Shukla (2020) Studied on the Image Restoration of Imagewith Gaussian Filter. In proposed
system Gaussian filter was been used and the varied reasonably noise was added hence de-blurring method was
employed to get a blurred image. When this image filtering was additionally enforced for removing these noises.
Tamilsel vi K and Prof. T. Thenmoz hi (2020) Restoration Techniques Available for Satellite Image Sensing
Applications – A Review. In this surveys various degradation models used for satellite images and presents the
inferences from the survey, which helps the readers to choose specific restorationfor specific degradation model a brief
detail on the various existing restoration techniques for satellite image restoration.
Hetvi Soni and Darshana Sankhe (2019) Image Restoration using Adaptive Median Filtering. The proposed work
adaptivemedian filtering was used to get a better image restoration as Generally median removes only the presence of
such noise but median filter can works fine for about intensity of 20% noise in the image.
Abhilash Bag (2018) A Review on Various Restoration Techniques in Digital Image Processing. The main objective of
there work was to carry out a comparative study. Though every technique may have got its own way of dealing with the
problem and have their own pros and cons. It also concluded from their explanations that usage of the techniques was
governed by the understanding, requirement and the standard of the output needed. They also said median filter works
better for impulse noise.
Basava Prasad B and Ravi M (2014) A Study On The Importance Of Image Processing And Its Applications. The goal of
their operation can be said to be divided into 3 categories. Firstly image processing in which input can be an image and
output was also an image; secondly image analysis where input may be an image and output are the dimensions or
measurements. Finally image understanding in which inputwas an image and output was in the standard description of
an image.
Shivani Dere, Anurag Chaudhari, Adarsh Laddha, Yashaswini Deora and Dhanalekshmi Yedurkar (2021)Digital
Image Colorization, Style Transfer And Image Inpainting using Deep Learning. The proposed method combines all the
three applications into a single web-based application termed as Neuron. Here, colorization is performedby CNN, image
inpainting is obtained by Generative Adversarial Network (GAN), style image is generated by Neural Style Transfer
(NST) techniques. We trained the distinct models for all three applications and produced qualitative and quantitative
comparisons with other traditional approaches to endorse this approch.
Sravya Vattem, Varun Kumar Bejugam, Janani Pokkuluri(2021) Style Transfer for Artistic Image Reconstruct ion
using Convolution al Neural Networks. Proposed system was implemented by generating the output image such that it
preserves some notions of the content imagewhile adapting to certain characteristics of the style image.
These characteristics were extracted from the images using aconvolutional neural network. The aim wasto implement a
loss function that will minimize the distance between the generated image and extracted content and style
representations.
3. METHODOLOGY
3.1 IMPLEMENTATION
We are collecting dataset and apply algorithm (CNN: - Convolutional Neural Net- works)Prepare a trained file to
compare with further peoples data. Convolutional Neural Networks is a popular deep learning technique for current

visual recognition tasks. There are four layered concepts in Convolutional Neural Networks:
• Convolution,
• ReLu,
• Pooling and
• Full Connectedness (Fully Connected Layer).
The input data features will be compared with the already trained image on the basis of closed match the result will be
predicted. Once the input image is processed the system further predicts whether the patient is infected with COVID-19
or Not. We have designed a system using python asbackend and HTML/CSS as front end, we have a webpage where we
can upload an image and submit and system thenprocess that image and output is predicted.
Fig-1: System Architecture
Fig-2: Sequence Diagram

Fig-3: Use Case Diagram
• Hidden layers consisting of convolution layers, ReLU (rectified linear unit) layers, the pooling layers, and a fully
connected Neural Network.
It is very important to understand that ANN or Artificial NeuralNetworks, made up of multiple neurons is not capable of
extracting features from the image. This is where a combination of convolution and pooling layers comes into the
picture. Similarly, the convolution and pooling layers can’t perform classification hence we need a fully connected Neural
Network. Before we jump into the concepts further let’s try and understand these individual segments separately.
The role of CNN is to reduce the images into a form that is easier to process, without losing features critical towards a
good prediction. This is important when we need to make thealgorithm scalable to massive datasets.
The challenge with images having multiple color channels is that we have huge volumes of data to work with which
makes the process computationally intensive. In other worlds think of it like a complicated process where the Neural
Network or anymachine learning algorithm has to work with three different data (R-G-B values in this case) to extract
features of the images and classify them into their appropriate categories.

machine learning algorithm that can take in an input image, assign importance (learnable weights and biases) to various
as- pects/objects in the image, and be able to differentiate onefrom the other.
CNN works by extracting features from the images. Any CNN
consists of the fol- lowing:
• The input layer which is a grayscale image
• The Output layer which is a binary or multi-class labels
3.3. Dataset
Dataset contain different type of images like Blur images,Low resolution images, High Resolution images which are used
for resolution.
4. CONCLUSIONS
We have proposed a single frame based SR approach which can adaptively choose parameter of regularization terms
while generating high space resolution image. To achieve self- adaptive parameter chosen, we also propose a robust
reference image quality assessment which focuses on blurring and ringing effect to provide feedback to regularization
terms. Our approach can effectively generate high resolution image from single input low resolution image using CNN
Techniques.
5. ACKNOWLEDGEMENT
We would like to thank our Principal Dr. S. D. Markande, Headof the Department Dr.Geeta S. Navale, our project Coordinator
Mrs. B. D. Shendkar, our guide Mrs. Asmita R. Kamble for theirvaluable advice and technical assistance.
6. REFERENCES
[1] Priyanka and Rishabh Shukla , “Image Restoration of Imagewith Gaussian Filter” , International Research Journal of
Engineering and Technology 2020.
[2] Tamilselvi. K and T. Thenmozhi, “Restoration Techniques Available for Satel- lite Image Sensing Applications – A
Review ”, International Research Journal of Engineering and Technology, 2020.
[3] Madhuri VM, Samyama Gunjal GH, Savitha Kamalapurkar , “Image Restora- tion using Adaptive Median Filtering”,
International Research Journal of En- gineering and Technology, 2019.
3.2 Algorithm Convolutional Neural Network (CNN)
CNN or the convolutional neural network (CNN) is a class of deep learning neural networks. In short think of CNN as a

[4] Mauro Castelli and Fabiana Martins Clemente, “A Review on Various Restora- tion Techniques in Digital mage
Processing” ,, International Research Journal of Engineering and Technology,2018.
[5] Basavaprasad B , Ravi M , “A Study On The Importance Of Image Processing And Its Applications ”, International
Journalof Research in Engineering and Technology , 2014.
[6] Shivani Dere, Anurag Chaudhari, Adarsh Laddha, Yashaswini Deora, Dhanalek- shmi Yedurkar , “Digital Image
Colorization, Style Transfer and Image In- painting using DeepLearning“ , International Research Journal of Engineering
andTechnology (IRJET) 2021.
[7] https://www.analyticsvidhya.com/blog/ 2021/05/deep-learning-for-image-super- resolution/
[8] https://developers.google.com/machine- learning/gan/ganstructure.
7. BIOGRAPHIES
Vinaya Akhade is currently a final year B. E. student of
savitribai phule pune University.
Rutuja Naikwade is currently a final year B. E.
student ofsavitribai phule pune University.
Harshada Sherkar is currently a final year B. E.
studentofsavitribai phule pune University.
Mrs. Asmita R. Kamble is currently working as a
Professor inSinhgad Institute of Technology and
Science, Narhe

Image super resolution using Generative Adversarial Network.

More Related Content

Similar to Image super resolution using Generative Adversarial Network.

More from IRJET Journal

Recently uploaded

Image super resolution using Generative Adversarial Network.