Loading...

Proposing a Method to Enhance the Quality of Image Data in the Data Mining Process

Mahmoudabadi, Batool | 2019

333 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53082 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Khedmati, Majid
  7. Abstract:
  8. Single image super-resolution (SISR) is a known difficult problem, which aims to obtain a high-resolution (HR) output from one of its low-resolution (LR) versions.The problem of Super Resolution has applications in many areas and industries where, for example, it can reduce the cost and time of re-imaging in many fields including the medical industry. In the satellite industry, distortions are eliminated and geographic information increases with clarity. In the astronomical industry, image recognition and computation become easier with super resolution. Also in many other important fields, for example, car license plate imaging, image quality enhancement can have significant effects for better detection of the related objects.To solve the SISR problem, recently powerful deep learning algorithms have been employed and achieved the state-of-the-art performance.The purpose of this study is to provide a model to improve the performance of neural networks in order to solve the problem of Super Resolution. In the presented network, several outputs created by the inner layers are used in the loss function. Increasing the network depth is thought to increase the network learning as the number of weights increases however, due to the long distance between the primary layers and the output, learning in these layers is reduced.This approach helps to create deeper networks, along with more learning, and alleviate the problems caused by deeper networks by utilizing the output of the middle layers in the target function. As a result, deeper network can allow more data to be used for training and Create a better performing network. Another advantage is to use the output created by the inner layers of the network to create a better output. Based on the results of the proposed method and comparing it with other methods, the network has provided a better performance for creating high resolution images
  9. Keywords:
  10. Super Resolution ; Convolutional Neural Network ; Image Processing ; Deep Learning ; Quality Enhancement ; Image Enhancement

 Digital Object List

 Bookmark

No TOC