SVD Based Noise Removal Technique: An Experimental Analysis

Main Article Content

Deepika Sharma
Pawanesh Abrol


Digital Images have been widely used in various high end applications such as computer vision, surveillance, forensics, industries etc. Noise is one of the prominent issues that affect the further analysis of images. There are different types of noise such as salt and pepper noise, gaussian noise, periodic noise, speckle noise etc. In order to remove noise from digital images, large numbers of methods have been proposed. Using various filtering methods such as median filtering, linear filtering, frequency domain methods, noise can be effectively reduced. However these different methods give mostly domain specific results. Recently another technique called Singular Value Decomposition (SVD) has been used for a variety of methods of image analysis and synthesis. It has been applied in areas like image compression, face recognition, audio video signalling, remote sensing, pattern recognition etc. In this research paper, a new SVD based technique for removing noise from digital images has been proposed. This model has been implemented and tested for the selected domain of digital images. Images have been taken from both generated database as well as from standard database. The experimental results have been presented and analysed to detect different varieties of noise. The analysis of results indicates that the proposed SVD model significantly removes noise from the input digital image set. This model may further be explored for authentication of the digital images.


Keywords: Image authenticity, Singular Value Decomposition, Noise Removal, filter, gray images.


Download data is not yet available.

Article Details