A variety of medical and satelite images are essential as sources of in sequence for study and understanding. When an image is transformed from one form to another such as scanning, transmitting, digitizing, storing etc., degradation occurs to the output image. For this reason, the output image wants to be better in order to be recovered . This paper presents a detailed survey on various noise detection and reduction algorithms. Different noises will affect the image in different ways. A detailed survey of the research is needed in order to design the filter which will execute the needed aspects along with handling most of the image filtering issues
Visualization is the best expression of our mind, so it is remarkable that images play a significant role in observation. Digital images are repeatedly infected with various types of noises. These irregularities arise due to a number of imperfections encountered in communication channels and image sensors. The effect of noise is to corrupt the image by replacing some of the original pixels from the image with new pixels whose luminance values are equal or near to the minimum or maximum of the allowable dynamic luminance range. The fundamental task of image processing is to remove noise from the digital images. The most frequently occurred types of noise are i) additive noise (e.g. Gaussian noise), ii) impulse noise, iii) multiplicative noise (e.g. speckle noise). Addition of noise in an image depends on how the image is created. If the image acquisition is direct in digital format, the mechanism i.e. CCD (Change Couple Device) detector can introduce noise. Other reasons may be electronic transmission of image data or if the image is scanned, the film grain is a source of noise. In Gaussian noise, the value of each pixel in the image is changed from its original value with a small amount. In order to recover the original image, the noise should be removed. To digitally process an image, it is first necessary to reduce the image to a series of numbers that can be manipulated by the computer. Each number representing the brightness value of the image at a particular location is called a picture element, or pixel. A typical digitized image may have 512 × 512 or roughly 250,000 pixels, although much larger images are becoming common. Once the image has been digitized, there are three basic operations that can be performed on it in the computer. For a point operation, a pixel value in the output image depends on a single pixel value in the input image. For local operations, several neighboring pixels in the input image determine the value of an output image pixel. In a global operation, all of the input image pixels contribute to an output image pixel value. These operations, taken singly or in combination, are the means by which the image is enhanced, restored, or compressed.
Image processing is an active area of research in such diverse fields as medicine, astronomy, microscopy, seismology, defense, industrial quality control, and the publication and entertainment industries. The concept of an image has expanded to include three-dimensional data sets (volume images), and even four-dimensional volume-time data sets. An example of the latter is a volume image of a beating heart, obtainable with x-ray, Computed Tomography (CT). PET, single-photon emission computed tomography (SPECT),
MRI, ultrasound, SAR, co focal microscopy, scanning tunneling microscopy, atomic force microscopy, and other modalities have been developed to provide digitized images directly. Digital images are widely available from the Internet, CD-ROMs, and inexpensive charge-coupled-device (CCD) cameras, scanners, and frame grabbers. Software for manipulating images is also widely available.
Additive noise is one of the most common problems in image processing. Even a very high resolution photo is bound to have some noise in it. For that photo a simple box blur may be sufficient, because even small features like eyelashes or cloth texture will be represented by a large group of pixels.
But unfortunately, this is not the case with video where realtime noise reduction is still a subject of many researches.
Salt and pepper noise is a form of noise usually seen on images. It uniquely represents itself as randomly occurring white and black pixels. An effective noise reduction approach for this type of noise involves the usage of a median filter or a contrast harmonic mean filter. Salt and pepper noise affects images in situations where the image is transferred quickly.
The speckle noise is commonly found in the ultrasound medical images. It is a granular noise. And it inherently exists in and decreases the quality of the Active Radar and Synthetic Aperture Radar (SAR) images. Speckle noise in conventional radar is caused from random fluctuations in the return signal from an object that is no bigger than a single image processing element. It boosts up the mean grey level of a local area. Speckle noise in SAR is usually more serious. And it causes difficulties for image interpretation. It is affected by coherent processing of backscattered signals.
Combining both Gaussian & Impulse Noise is called Mixed Noise.
Poisson noise has a probability density function of a Poisson distribution.
In 1993 [1] FarhadEsfahani& Mark Richardson used Median Filter with Linear Filter to remove Impulsive noise from image. These techniques had good minimization error. But the computation time was very high.
In 1996 [2] Eduardo Abreu, Michael lightstone have proposed a novel and efficient approach for removing impulse noise from highly corrupted images using Rank order mean filter. The technique achieves an excellent tradeoff between noise suppression and detailed preservation with little increase in computational complexity over the simple median filter.
In 1997 [3] robust filtering method based on fuzzy logic was proposed by Young Sik Choi and Raghu to evaluate the situation at a particular pixel. We used the total compatibility of the neighboring pixels with the center pixel. The total compatibility is the mean value of the membership degrees to which the neighboring pixels represent a center pixel in a given window. Each filter and the corresponding condition in which it is to be applied constitute a fuzzy if–then rule in the proposed fuzzy rulebased system for image enhancement. When images are severely contaminated and can be used for smoothing noise while preserving edges, it removes the gray image noise only and the computational time is high.
This drawback was overcome by Harish Kundra , Monika Verma & Aashima in 2002 [4] when they denoised Salt & Pepper noise from image of TajMahal in 15 sec using 8- neighborhood method & also preserved the intricate features of the image.
In October 2002 [5] Michael Elad proposed Bilateral Filter. This method used anisotropic diffusion (AD), weighted least squares (WLS), and robust estimation (RE) for noise detection and reduction. It improves the accuracy and processing speed. It removes additive noise only.
In similar way Dimitri Van De Ville, Mike Nachtegael, Dietrich Van der we ken in 2003 [6] removed additive noise from Camera-man & Boat. This Filter was very feasible, fast, and simple & enabled fast hardware implementation.
In 2004 [7] ZHANG,xiao- guang”, XU jian- Jian',LI yu' used Fuzzy Neural network model to remove Gaussian noise from image. These techniques improve accuracy. But it affects fuzzy membership function.
At the end of 2005 [8] Chang-Shing Lee, Shu-Mei Guo, and Chin-Yuan Hsu introduced a filter which used Image Knowledge Base for noise detection & Parallel Fuzzy Inference, Fuzzy mean process & Fuzzy decision process for impulse noise removal on the image of Lena, Albert, Baboon, Cameraman, Sailboat, Bridge, Boats, House, Pentagon. It has High quality of global restoration, but did not handle Mix impulsive noise model and GIF for color image
Chang-Shing Lee, Shu-Mei Guo, and Chin-Yuan Hsu have proposed Genetic-Based Fuzzy Image Filter and Its Application to Image Processing in Aug 2005 [9], Image Knowledge Based approach for using noise detection processing, Parallel Fuzzy mean filter for noise reduction. It is suitable for Mix impulsive noise model only. This Genetic- Based approach gives better Noise detection performance on images.
In November 2005 [10] Roman Garnett, Timothy Huegerich, Charles Chui introduced a Universal Noise Removal Algorithm with an Impulse Detector. This algorithm is used for identifying noisy pixels. The resulting trilateral filter performs well in removing Gaussian and mixed noise as well as in removing impulse noise. It is not suitable for high density noisy images.
Some months later in Aug 2006 [11] HamedVahdatNejad, Hameed Reza Pourreza, and HasanEbrahimi introduced feasible filter. where for detection ,they used Fuzzy rule base system which associates a degree to each pixel & for removal, they used Weighting Contribution of neighboring pixel for removing Gaussian noise from Cameraman & Lena .
In Feb 2007 [12] Nguyen Minh Thanh and Mu-Song Chen have proposed Image Denoising Using Adaptive Neuro- Fuzzy System. He proposed Combination of multilayer Neuro-fuzzy structure model for Noise detection. It removes Impulse noise and noise detection performance is well. But it does not remove other kind of noises.
In October 2007 [13] Stefan Schulte, Samuel Morillas, Valentín Gregori, and Etienne E. Kerre have proposed a New Fuzzy Color Correlated Impulse Noise Reduction Method. It is a Vector based approach method for using noise detection. It reduces all kinds of impulse noise (for low and high noise levels) effectively. It preserves edge sharpness, and it does not introduce blurring artifacts or new colors artifacts in comparison to other state-of-the art methods. This method also illustrates that color images should be treated differently than grayscale images in order to increase the visual performance. As future work, we will extend the discussed filter to reduce also –stable (a mixture of Gaussian and impulse noise) very efficient, i.e., we will investigate the incorporation of an additive noise reduction method.
Buyue Zhang, and Jan P. Allebach have proposed Adaptive Bilateral Filter for Sharpness Enhancement and Noise Removal in May 2008 [14].The Adaptive Bilateral Filter (ABF) outperforms the bilateral filter in noise removal. At the same time, it renders much sharper images than the bilateral filter does. As a result, the overall quality of the restored image is significantly improved. Second, the ABF does not perform as well at corners as it does on lines and spatially slow-varying curves, since the ABF is primarily based on transforming the histogram of the local data, which cannot effectively represent 2-D structures. [14]
In Aug 2008 [15] M. T¨ulinYıldırım, AlperBasturkAnd M. EminY¨uksel used Type-2 Neuro-fuzzy operator (where membership function are also fuzzy) and then postprocessor for detection & for removal, Neuro-Fuzzy filter having Type-2 Neuro-fuzzy filter &Defuzzifier. They had removed impulsive noise from Baboon, Boats, and Bridge & Pentagon. And Preserved thin line edges & texture & other useful information in images
In June 2009 [16] Yu-Mei Huang, Michael K. Ng, and You- Wei Wen have proposed Fast Image Restoration Methods for Impulse and Gaussian Noises Removal. Noisy image pixels where these are detected by median-type filters. It is a Minimization algorithm for using noise detection technique.This algorithm is very efficient and the quality of restored images is good. It contains slow process.
In 2009 [17] K. Than gavel,R. Manavalan, I. Laurence Aroquiaraj used Statistical Method and Special Filters to remove Speckle Noise from image. The noise removal performance in this technique is high, but it removes only speckle noise.
In July 2009, Samuel et al., [18] have proposed a noise detection method base fuzzy peer group. This method removes Gaussian, impulse and mixed noises using fuzzy average filter and vector median filter. It removes the noise and maintains the quality of the image. But it takes more time for detection and removal of noise.
In 2010 [19] Jagadish H. Pujar used Fuzzy noise detection method & Fuzzy median filter for dealing with Salt & Pepper noise in Lena which preserve image detail very well, its smoothing rate is also preponderant.
In September 2010, Chih-Hsing Lin et al., put forward a novel switching filter named sorted quadrant median vector (SQMV) [20]. It also removes all types of noise efficiently, but it does not preserve edge detail.
In March 2011, Floran et al., proposed a filter bank which removes Gaussian and mixed poisson noises [21]. But it uses color image components only.
In 2011 [22] ShamikTiwari, Ajay Kumar Singh & V.P. Shukla used Statistical feature of noise type &Feedforwad Back Propagation Neural Network for removing Salt & Pepper, Gaussian and Speckle noise. It has showed better identification of noise using prelevantly used image filter. But the success is based on the training.
In April 2011, Tom Melange et al., have proposed a nonlinear filter to random value impulse noise [23]. It performs better than other methods, but it removes impulse noise only and only color images are applicable.
In July 2011, Aborisode D. O has proposed a filter based on fuzzy logic to remove impulse noise [24]. This method detects the noise in all locations and replaces the values on affected locations only. But this method also removes impulse only.
In Feb 2012 [25] R.Pushpavalli, G.Sivaradje have proposed a noise detection method. This method uses Switching median method for Comparing max, min value noise Decision Mechanism. This paper deal with high density impulse noise only.
In Aug 2012 [26] Sukomal Mehta, SanjeevDhull have proposed another noise detection method. This paper contains Detect Noisy Pixels Using fuzzy reasoning with lowest uncertainty. This Adaptive filter removes the Impulse Noise. It considers the high level of noise. The adaptive algorithm performed quite well.
In 2012, Punyaban Patel et. al., have proposed a noise detection and reduction method using fuzzy logic [27] . This method designed a fuzzy based adaptive mean filter to remove impulse, Gaussian and speckle noise. It removes all these noises efficienty, but it works for gray image only.
In 2012, Aheret. et.al., have proposed a switching scheme for noise detection and genetic algorithm for reduction [28]. This method used a supervised learning algorithm using non-linear filters. It removes impulse and Gaussian noise for grayscale image. The computation time is high for this method.
In April 2012, Bo Xiong et al., have proposed a novel algorithm for noise detection and reduction [29]. The novel NL-means noise detection and reduction algorithm removes impulse noise and achieves higher PSNR values. But it removes impulse noise only.
In February 2013, Hsien-Hsin Chou etal., have proposed a parallel fuzzy inference mechanism based fuzzy image filters to remove impulse noise [30]. This algorithm is optimal, but it does not filter other noises.
In March 2013, IyadF.Jafar et al have proposed a switching median filter for boundary discriminative noise detection [31]. This filter removes high density noise in the image and expands the filtering window. But this method also works for impulse noise only.
In April 2013, HossineTalebi et al., have proposed a transform domain filter to remove mixed noise [32]. This method works faster but the computational cost is high.
In October 2013, Joan-Gerard Camarena et al., have proposed a noise detection method named fuzzy filter and vector median filter to remove Gaussian, impulse and mixed noises [33]. This method performs better than other methods, but the computational time is high.
A variety of surveys have been done in this paper. We have discussed various denoising algorithms and techniques.
From the literature, we have analyzed that fuzzy and genetic based filters perform well. It is also clear that many researches are going into noise detection and removal. Threre are many applications in this technology. Hence, there is a need to design a noise detector and removal which should detect and remove all types of noise such as Gaussian, Impulse, and Mixed noise in less computational time and cost.
It should also increase the quality of the image. For noise detection, fuzzy based technique can be used, as that technique performs well. For noise removal, adaptive filters are suggested which achieve better PSNR (Peak signal – to – noise ratio) than other filters. A combined approach of these methods may achieve higher PSNR in less computational time and cost.
Image Filtering plays a vital role in many applications. This paper provides a detailed study of image filtering for removing Salt & pepper, Gaussian and Speckle noise from the Gray and color Images and also which type of noise detection and reduction techniques applied for all filters has been explained in a clear maner. It provides us the track of how the issues were faced while designing a filter and when, how and by whom it was resolved. This study helps to choose the proper filter for removing noise from images