International Journal of Computer Applications (0975 – 8887) Volume 44– No.7, April 2012 19 Asymmetric Trimmed Median Filter for Images Highly Corrupted with Random valued Impulse Noise R.Pushpavalli Electronics and Communication Engineering Pondicherry Engineering College Puducherry, India-605 014 G.Sivaradje Electronics and Communication Engineering Pondicherry Engineering College Puducherry, India-605 014 ABSTRACT Asymmetric Trimmed Median Filter for Image denoising is proposed in this paper. This technique can be used for restoring the images extremely corrupted with random valued impulse noise. This paper introduces an impulse detection technique and decision based median filter for restoring the corrupted images. The detection technique is used for discriminating between corrupted and uncorrupted image pixels. The corrupted pixels are restored using Asymmetric trimmed median filter. The performance of the proposed restoring scheme is evaluated with random valued impulse noise for different test images. Simulation results show that this method is significantly better than a number of existing techniques in terms of image restoration and noise detection. Keywords Impulse noise, Median filters, Image processing, Restoration. 1. INTRODUCTION Digital images are often corrupted by impulse noise due to transmission errors, malfunctioning pixel elements in the camera sensors, faulty memory locations, and timing errors in analog-to-digital conversion. In most applications, denoising the image is fundamental to subsequent image processing operations, such as edge detection, image segmentation, object recognition, etc [1-3]. The goal of noise removal is to suppress the noise while preserving image details. Removal of the impulse noise is done in two stages: detection of noisy pixel and replacement of that pixel. Median filter is used as a backbone for removal of impulse noise [4]. Many filters with an impulse detector have been proposed to remove impulse noise. One of the most popular methods is the median filter, which can suppress noise with high computational efficiency. Since every pixel in the image is replaced by the median value in its neighborhood, the median filter often removes desirable details within the filtering window on the image and blurs it too. Several impulse noise removal methods with different kinds of noise detectors have been proposed in the literature, namely; switching based median filter, weighted median filter, center weighted median filter, Decision based median filters, etc [5-7]. The main drawback of these filters is that median values or their variations are used to restore the noisy pixels, and hence these median based filters are alters both noise free and noisy pixels. As a result, this performance will prone to misclassify the pixel’s characteristics. Some switching-based median filtering methodologies had been proposed for impulse noise elimination [8-22]. These filtering operations were obtained by applying “no filtering” to preserve true pixels and standard median filter to remove impulse noise. The median based filtering operations are crucial to achieve good filtering performance, especially at high noise density interference. Removal of random valued impulse noise algorithms have been developed [23-26]. Removal of random valued impulse noise from images without losing their features such as edges and fine details is difficult tasks in image filtering. In order to address these issues, in this paper a powerful image restoring technique for highly corrupted images, namely Asymmetric trimmed median filter for images highly corrupted with random valued impulse noise is proposed. This filter is obtained in two stages; Noisy pixels of the corrupted image are identified using a Impulse Detection Technique in first stage. In this stage, the corrupted and uncorrupted pixels in the image are detected by checking the pixel element value against the dynamic range of high noise level (HNL) and low noise level (LNL) respectively. These values are the impulse noise intensity values. It is followed by an Asymmetric trimmed median filter (ATM) for recovering those corrupted pixels identified from IDT in the second stage. In an impulse detection stage, the current pixel is detected as an uncorrupted pixel and it is left unaltered, otherwise, it is corrupted. Then the proposed filtering technique is performed on it. The proposed filtering technique of impulse detection followed by ATM filter is capable of producing high quality images and it prevents image blurring compared to other denoising techniques. Extensive simulation experiments have been conducted to evaluate and determine the performance of the proposed filter in terms of quantitative and qualitative metrics. The outline of this paper is as follows. In Section II, define the noise model. Section III describes noise detection and filtering technique in detail. Section IV gives simulation results to demonstrate the performance of the proposed filtering technique. Finally conclusions are drawn in section V. 2. NOISE MODEL Impulse noise is caused by malfunctioning pixels in camera sensors, faulty memory locations in hardware, erroneous transmission in a channel. Generally two types of types of impulse nose namely Salt & Pepper and random valued impulse noise [1] respectively. Salt & pepper noise represents the pixel value of maximum (255) and minimum (0) intensity on digital images. Whereas, in the case of random valued impulse, digital images are often corrupted by any random value in the dynamic range of grayscale. The proposed filter only detects random valued impulse noise present in digital images in very efficient manner then removes it. As the impulse noise is additive in nature, noise present in a region does not depend upon the intensities of pixels in that region. Based on the properties of probability, images are corrupted with impulse noise. The intensity of grayscale pixel