Improved Switching-Basedmedian Filter For Impulse Noise Removal

This thesis proposed a new algorithm to reduce impulse noise from digital images. In order to achieve this, thorough literature surveys on impulse noise models and median filtering frameworks have been carried out successfully. The proposed algorithm is based on switching median filtering appr...

Full description

Saved in:
Bibliographic Details
Main Author: Teoh, Sin Hoong
Format: Thesis
Language:English
Published: 2013
Subjects:
Online Access:http://eprints.usm.my/44005/1/Teoh%20Sin%20Hoong24.pdf
http://eprints.usm.my/44005/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This thesis proposed a new algorithm to reduce impulse noise from digital images. In order to achieve this, thorough literature surveys on impulse noise models and median filtering frameworks have been carried out successfully. The proposed algorithm is based on switching median filtering approaches. The method can be generally divided into two main stages, which are impulse noise detection stage and impulse noise cancellation stage. Modifications towards a well known boundary discriminative detection (BDND) method have been made. First, rather than using any sorting algorithm, the local median values were determined from manipulated local histograms. Next, in the noise detection stage, in addition to the originally proposed intensity distance differential approach, the new method includes intensity height differential approach to reduce false detection rate. Then, instead of using adaptive approach for noise cancellation stage, the proposed method utilizes iterative approach. Broad impulse noise model has been employed for the evaluation process, to investigate the robustness of the method. Based on the evaluations from root mean square error (RMSE), false positive detection rate, false negative detection rate, mean structure similarity index (MSSIM), processing time, and visual inspection, it is shown that the proposed method is the best method when compared with seven other state-of-the art median filtering methods.