複合雜音 除去를 위한2D 多重 디지털 필터에 관한 硏究

Abstract

Abstract

The recent information communication makes transmissions by using a variety of media and also the digital devices related to hardware and software are being popularized. Since the time when the ubiquitous world was mentioned, TV, computer, smart phone, etc. have played an important role in the image content industry as the most prominent media to use image signals. Moreover, they can be deemed as a large axis to form the modern society. However, noises are added to image due to various causes such as external environment conditions and interference of an internal system in the course of image pre-processing, image data transmission and storage, and image degradation phenomenon occurs consequently. Therefore, there is a growing awareness of the importance of noise removal in order to recover image degradation, and the studies related hereto are being actively conducted. The noise added to image can be classified into the three types of noise depending on the occurrence cause and type. These are as follows: impulse noise, AWGN (additive white Gaussian noise) and composite noise in which the two noises are overlapped. One of the most prominent task as for most of image processing processes is to remove noise. On this account, many techniques to remove such noise have been proposed and they include linear filter and non-linear filter mainly in spatial areas. In general, the linear filters include MF (mean filter), GF (Gaussian filter), etc. These are relatively simple and effective in low-frequency range. Also, they are visually soft. However, they have the blurry phenomenon in an area in which the gray level of an image is rapidly changed. The non-linear filters include SMF (standard median filter), A-TMF (alpha trimmed mean filter), CWMF (center weighted median filter), ACWMF (adaptive center weighted median filter), SAWM (self-adaptive weighted filter), etc. They are excellent for removing noise when high-density noise is added. Also, they have an outstanding feature of energy retention. However, these filters have the problem of deforming the original pixels or distorting the outline since they generally filter both of noise pixels and non-noise pixels when removing noise. This thesis proposed the two dimension multi-digital filter to process impulse noise and AWGN after separating them in accordance with the noise type in a spatial area in order to remove composite noise that is added to an image. The proposed filter first detects impulse noise. When it was impulse noise, a non-linear filter was designed to obtain an excellent energy retention feature by sub-dividing the mask into 4 areas. When it was non-impulse noise, it was the case in which pixels were damaged by AWGN. It was designed by configuring a strong filter in parallel in low and high frequency. This filter was configured by applying the spatial weighted value that applied the spatial distance difference between the central pixel and surrounding pixel inside the mask, the adaptive weighted value that applied the central pixel value and surrounding pixel value and the adaptive weighted value by the critical value of difference between the surrounding pixel and pixel value. Therefore, the proposed algorithm represents an excellent noise removal feature when removing composite noise and they also showed an outstanding result in low and high frequency ranges. Also, the existing methods were compared in order to confirm the excellence of proposed algorithm.