High Density Impulse Noise Removal and Edge Detection in SAR Images based on Frequency and Spatial Domain Filtering
S.Ranjitha1, S G Hiremath2

1S.Ranjitha, Research scholar, VTU, Department of ECE, East West Institute of Technology, Bengaluru (Karnataka), India.
2Dr. S G Hiremath, Professor and Head, Department of ECE, East West Institute of Technology, Bengaluru (Karnataka), India.

Manuscript received on 18 February 2019 | Revised Manuscript received on 27 February 2019 | Manuscript published on 28 February 2019 | PP: 643-648 | Volume-8 Issue-3, February 2019 | Retrieval Number: C5996028319/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Development in the science and technology, has been a great improvement in the field of space technology. Now a day’s we can even capture the image from the different layers of the earth. This can be referred as Synthetic Aperture Radar (SAR) imaging. As this distance between lens and object increases, it becomes difficult to get the clear and noise-free image. There are many factors which degrade the image in different ways. One such degradation can be in the form of salt and pepper noise. This constitutes for presence of white and black spots on the image. So it is necessary to remove this noise, and to obtain a much clearer image. By taking the advantage of both spatial domain and frequency domain filter, a more effective method of de-noising is proposed. The image is denoised in three folds. First step includes preprocessing by using spatial domain filters, second stage uses frequency domain filter to avoid blurring and smoothing effect on the image. In this stage we use different methods to separate noisy and noiseless pixels, such as any machine learning or deep learning method (ANN, CNN, SVM). Thus maintain the textural information. Last stage uses spatial domain filter to remove any residual noise present. Thus obtained image resembles more with the original image.
Keywords: Artificial Neural Network (ANN), Support Vector Machine (SVM), Convolutional Neural Network (CNN), Batch Normalization (BN), Rectified Linear Unit (ReLU)

Scope of the Article: Artificial Neural Network