복합잡음 환경에서의 비선형필터

Abstract

A Nonlinear Filter in Complex Noise Environment.

Kyung Hyo Lee

Department of Control and Instrumentation Engineering Graduate School of Pukyong National University

Abstract

With the development of the information technology in recent years, the innovation of multimedia information technology also has been accelerated. Former times information technology generally has only depended on text or sound, while nowadays information is being moved through a variety of image media. Cell phone, TV and computer have been major elements of modern society as mediators using image signal. Therefore, image signal processing also has been treated importantly and done actively. The processing has been developed in many fields of digital image processing technologies as image data compression, recognition, restoration, etc. Noises are inevitably generated by using the signals during the processing, and typical types of the noise are impulse(salt & pepper) and AWGN(addiction white gaussian noise). To reduce the noise, various

- ⅱ -

environment, SM(standard median) filter is simple and it shows a kinds of filters have been developed, and according to each noise, it is being used different filter each. The image filter used digital image process basically has a two-dimensional structure. There are two methods of the filter's creation - One is reiterating one dimension, and the other is using an indivisible two-dimension image filter. The space domain method using mask is the latter one. That is inserting the value-multiplied pixels values faced each other when the two-dimension filter overlapped on input image- to the filter value's center position and the same position in the image. The image filter is being used widely along with one-dimension filter, according each noise. Most people are using various filters to remove the impulse noise. The mean filter is typical linear filter and blurring filter. The blurring is a effective filter to removal of Gaussian noise but ineffective filter to removal of impulse Noise. Because the min-max filter uses removal of peak value, it have good result to removal of impulse - ise but many errors. The median filter is not only conservation of edge but also removal of impulse noise. But it is ineffective to removal of Gaussian noise. So, the noise is not generated by one signal but by a complex. In this paper, I suggested an image filter to remove the complex noise, and compared with existing filters’methods for verification.