Comparison of LULU and Median Filter for Image Denoising

Abstract

Abstract-This paper presents a comparison between LULU and Median filters for impulse noise in images. Noise removal from images is always a challenging area of research. Different methods are being used for different image noises such as Wiener filter for Gaussian noise, Frost filter for speckle noise and median filter for impulse noise. LULU filters are widely being used for impulse noise too. LULU filters are nonlinear rank selector operators which are computationally more competent and the performance of the operator is straightforward to describe. The comparison between Median and LULU filters, in this paper, is performed using two different image quality metrics which are RMSE and PSNR.Index Terms-LULU filters, impulse noise, RMSE, PSNR. I. INTRODUCTIONIn the field of image processing, there are so many issues and complexities for the recovery of original data from the noise corrupted data. In general, image noise is the variation of color or brightness intensity which can be caused by different sources. In general, image sensors degrade the quality of the images. Faulty devices, troubles with the data collection procedure, and interfering natural cause can all corrupt the data. Moreover, noise can be caused by compression and transmission errors. Thus, noise removal is required and is the first step before images are analyzed. It is necessary to apply an efficient noise reduction method to compensate for such data corruption [1], [2]. Even though there are so many different methodologies in this field for noise removal, image denoising is still a big issue. Some noise filters cause blurring of the images, because in reality the nature of the noise is unknown or only known to some extent.Impulse noise results due to the malfunctioning of the image sensors and the transmission channels. It can be identified by black and white points in the image. So far, there are so many different algorithms that had been proposed to filter this type of noise such as median, switching median, weighted median, adaptive median, rank conditioned median, separable median and contra harmonic mean filter. There are different names for impulse noise which are salt and pepper as well as shot noise.In this work, we address impulse noise. Generally, median filter is used for reducing impulse noise. We compared four different LULU filters with median filter. This comparison Manuscript received April 20, 2013; revised June 28, 2013 It is extensively applied in digital image processing due to its property of protecting edges as well as at the same time eliminating noise but it causes the image to blur. But still Median filter causes significantly a smaller amount of blurring than linear smooth filters of the same size. In general, all Median filters produce reasonable results for both bipolar and unipolar impulse noise. They are nonlinear, non-idempotent smoother which is computationally less efficient than LULU. A. 1D MedianIn one dimensional arrays, median filter is given by:The Fig. 2 illustrates the median operator in filtering a s...

Image Enhancement and De-Noising Techniques of Magnetic Resonance Images

“…The output is set to this median value with the coordinates of pixel (m,n) and without the smoothing characteristics. The edges and image intensity are not affected in the magnetic resonance images with respect to their gray level intensity [47]. This nonlinear filtering method is used to remove the shot noise from the image without attenuation and median window slides pixels are processed without intensity [48].…”

MRI Medical Image Denoising by Fundamental Filters

“…Nowadays, MR image de-noising has become an important purpose in medical imaging particularly the Magnetic Resonance Imaging (MRI). Many de-noising and enhancement techniques are applied on MRI images [1][2][3][4][5][6][7][8].…”

Multispectral and thermal images for acne vulgaris classification

“…We concluded that Image processing is one state of the art tool for introducing objectivity and automation in skin lesion diagnosis. Through image processing techniques, one can segment the lesion from normal skin, can categorize lesions based on features, remove artifacts, and estimate thickness of the lesion through 3D reconstruction [12]. In this way, almost all diagnostic functions that a human can do are also performable by image processing techniques in an objective way.…”