Face Image Super Resolution using Machine Learning

Project Title

Face Image Super Resolution using Machine Learning

Project Area of Specialization Artificial IntelligenceProject Summary

Image Super-Resolution (SR) is an important class of image processing techniques to enhance the resolution of images and videos in computer vision from low-resolution (LR) to high (HR). Recent years have witnessed remarkable progress of image super-resolution using deep learning techniques. Image resolution describes the details contained in an image. The higher the resolution, and the more the image details. Super-resolution (SR) are techniques that construct high-resolution (HR) images from several observed images, thereby increasing the high frequency components and removing the degradations caused by the imaging process of the low-resolution camera. The basic idea behind Super Resolution is to combine the non-redundant information contained in low-resolution frames to generate a high-resolution image. Our primary purpose is to train our system in order to pre-process and post process the images from the collected dataset. The goal of our proposed application is to use the concept of machine learning along with CNN and Super Resolution (SR) and to apply it to train our application in order to convert the lowresolution image into the high-resolution image. The approach teaches end-to-end mapping directly between low and high-resolution images. This application will be used for surveillance to detect, identify, and perform facial recognition on low-resolution images obtained from security cameras.

Project Objectives

?  To create an application that takes a low resolution image as an input and converts it into a high resolution image. 
?  To make the application predict the actual image accurately using machine learning technique.

Project Implementation Method

a. CNN 

Convolutional Neural Network (CNN) is trained to encode high level information about the class of images being imaged; this information is utilized to mitigate artifacts in intermediate images produced by use of an iterative method. It is a class of deep neural networks, most commonly applied to analyzing visual imagery and uses relatively little pre-processing compared to other image classification algorithms. This means that the network learns the filters that in traditional algorithms were hand-engineered. This independence from prior knowledge and human effort in feature design is a major advantage.  

b. Depth Maps 

Depth map is a kind of image which is composed of the gray pixels defined by 0 ~ 255 values. The 0 value of gray pixels stand for that 3D pixels are located at the most distant place in the 3D scene while the 255 value of gray pixels stand for that 3D pixels are located at the most near place. In depth map, each depth pixel would define the position in Z-axis where its corresponding 2D pixel will be located. Since it can be referred as pixel-by-pixel to produce realistic 3D image 

c. Super Resolution Techniques 

Sr techniques are used to obtain a high-resolution (HR) image (or sequence) from observed multiple low-resolution (LR) images. The SR image reconstruction is proved to be useful in many practical cases where multiple frames of the same scene can be obtained. Which will help in producing better results.  

d. Deep learning 

Deep learning denoising network outperforms traditional algorithms in Poisson denoising especially when the noise is strong which generates clearer images as results. 

e. Generative Adversarial Nets (GAN) 

A variety of deep learning methods have been applied to tackle SR tasks, ranging from the early Convolutional Neural Networks (CNN) based method to recent promising SR approaches using Generative Adversarial Nets (GAN). A Generative Adversarial Network, or GAN, is a type of neural network architecture for generative modeling which involves using a model to generate new examples that plausibly come from an existing distribution of samples, such as generating new photographs that are similar but specifically different from a dataset of existing photographs.

Benefits of the Project

Most of the times we observe that face images obtained by an outdoor surveillance camera, are often confronted with severe degradations (e.g., low-resolution, low-contrast, and noise). This significantly limits the performance of face recognition systems. Therefore, our objective is to make an SR based application that will be very useful in this concern and can be used in such cases. It will take low resolution static images as input and convert them into high resolution images, that is, it will produce the actual images as the output. Our aim is to attempt a combination of partial image restoration using SR. It has several other practical applications such as medical imaging, surveillance and security.

Technical Details of Final Deliverable

The final deliverable will be a software application which takes a low resolution image as input and converts it into high resolution image as output . Moreover, this application will be able to predict and reconstruct the actual image using GANs and various other machine learning algorithms.

Final Deliverable of the Project Software SystemCore Industry ITOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Decent Work and Economic GrowthRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	63800
Web hosting	Equipment	12	2400	28800
GeForce GTX GPU	Equipment	1	34000	34000
Report printing	Miscellaneous	1	1000	1000