Noise Removal and Filtering Techniques used in Medical Images

Noise removal techniques have become an essential practice in medical imaging application for the study of anatomical structure and image processing of MRI medical images. To report these issues many de-noising algorithm has been developed like Weiner filter, Gaussian filter, median filter etc. In this research work is done with only three of the above filters which are already mentioned were successfully used in medical imaging. The most commonly affected noises in medical MRI image are Salt and Pepper, Speckle, Gaussian and Poisson noise. The medical images taken for comparison include MRI images, in gray scale and RGB. The performances of these algorithms are examined for various noise types which are salt-and-pepper, Poisson, speckle, blurred and Gaussian Noise. The evaluation of these algorithms is done by the measures of the image file size, histogram and clarity scale of the images. The median filter performs better for removing salt-and-pepper noise and Poisson Noise for images in gray scale, and Weiner filter performs better for removing Speckle and Gaussian Noise and Gaussian filter for the Blurred Noise as suggested in the experimental results.


INTRODUCTION
N oise is caused due to various sources which include many external causes in transmission system and environmental factors which includes noise like Gaussian, Poisson, Blurred , Speckle and salt-and-pepper noise.Noise removing method has become an important factor in medical imaging applications and the most commonly used filters are Median filter, Gaussian filter, Weinerfilter which gives the best result for the respective noises.
The need for the smoothening of images has becomes essential whichis required to remove the noise and for that best filters or standard filters are used in most of the image processing applications.The property of a de-noising model is to remove the noise from the image and also preserve the edges.There are two types of models which are used for de-noising linear model and non-liner model.Most of the time linear models are experimented because of its speed even though it has the limitation of not able to preserve the edges of an image in a efficient way.These data is observed by using filters and finding out the best filter on the basis of the histogram, size and clarity of the MRI images given to these filters.

Removal Techniques
Image de-noising is assetfor image noise processing which includes filtering techniques which includes different ways to de-noise an image.It is solved by using different algorithms.Accordingly, noises are spotted with neighboring information and are removed using best filtering techniques without affecting the image quality and reinforce the smoothness of the image taken for examination.

Median Filter
The Median filter is nonlinear technique which is known as order-statistic filtering in digital image processing.Median filter is very popular technique for the removal of impulse noisebecause it runs through the signal cell by cell and replacing the value of each cell with the neighboring by a median of the intensity levels with its mathematical accuracy.The outcome of neighborhood pixels by the Median Filter in an image is done by the static filtering window size which slides cell by cell over the signal.The technique is applied across the image and therefore ittends to transform both noisy and de-noised pixels present in the image.Due to this tendency of median filter the good pixels cell are replaced by the corrupted ones.Therefore, de-noising often leads to removal of fine details present in an image because it is done at the cost of distorted and blurred features possessed by the this filtering technique.

Weiner Filter (WF)
The goal of the Weiner filter is tofilter out the noisepresent in an image which posses corrupted signal in it.This filtering technique uses statistical approach to filter the noise from each pixel of an image.This filteringtechnique uses different angle in an image to modify the corrupted signal in it.Original image signal has spectral properties and noise present in it so to start with experiment one should have the knowledge of the properties of it, one seeks the LTI filter (Linearity and Time-Invariance) whose outcome will be closer to the original signal present in the image as achievable.Wiener filter is a technique which performs optimal trading involving opposite filtering and noise smoothing.It removes the blurring and additional noise present in the image and it is also very optimal in relation to the mean squared error where it minimizes the overall Mean Square Error in the operation of the filtering technique for noise removal 1 .
Wiener filters are usually defined by the following: a.
hypothesis: additive noise and image signal are inactive linear random processes containing spectral characteristics.b.
Necessity: The filter must be able to achieve and can be accessed.c.
Performance criteria: It depends on minimum Mean Square Error.

Gaussian filter
Speckle Noise is typical noises which is caused due to internal or external factor and are generally present in the digital images and MRI images.Gaussian filter is implemented to remove the Speckle Noise present in ultra sound images or MRI brain images.In this technique, the average value of the surrounding pixel or neighboring pixels replaces the noisy pixel present in the image which is based on Gaussian distribution.

Different type of noise in Medical images
The process which attempt to remove the noise from the image and restore the quality of the original image is known as Image Restoration.This is an important aspect in maintaining the quality of the image by restoring the pixel value.Restoration techniques area model for linear image degradation and it isthe opposite process to improve the quality of original image.To obtain an optimal estimate of the desired result restoration technique involves mathematically principle of goodness which helps to achieve.

Gaussian Noise
Gaussian distribution which is also known as normal distribution whose Probability Density Function is equal to statistical noise known as Gaussian Noise.This noise is removed from the digital images by smoothening of the image pixels which helps in reducing the intensity of the noise present in the image which is caused due to acquisition but the result may be sometime undesirable and also which can result in blurring edges of the high-quality images 2 .
The formula of adding the Gaussian Noise to an image is: g = imnoise (I,'Gaussian', m, var), where I is the input image, m is mean and var is variance.

Salt Pepper Noise
The image which is low in quality has bright and dark pixels present in it which causes noise in it also referred as Salt Pepper noise.This noise will generally have bright pixels in dark portion and dark pixels in bright portion of the image.Black and white dots appear in the image 3 as a result of this noise shown in the fig 10(a).Due to sharp and unexpected changes of image signal the noise arises and causes dead pixels, analog-to-digital converter errors, etc. in the image.This kind of noise can be removed by using Dark Frame Subtraction (DFS) and by constructing new data points around dark and bright pixels which is obtained by the Median filter or morphological filter 4 .

Speckle Noise
The Speckle Noise is defined as a noise which is present in the images and which degrades the quality of an image.Speckle Noise is a incident that convoys all rational imaging model quality in which images are formed by inquisitive echoes of a mediate waveform that originate from diversity of the studied objects 5 .These are the granular noises that are fundamentally present in the image and reduce the quality of the active radar and Synthetic Aper ture Radar (SAR) images or Magnetic Resonance 6 .Imaging (MRI) images is referred to as Speckle Noise.If Speckle Noiseis present in the images then it results in the random variations ofthe return signal which increases the grey level in an image.A Speckle Noise is the coherent imaging of objects in the image.In fact, it is caused due to errors in data transmission.This kind of noise affects the ultrasound images and MRI images.Speckle Noise follows a gamma distribution and is given as: (g) = [5ØTÜ5ØTÜ"""1 ("""1)!5Øþ<5ØNÜ"" e"5ØTÜ5ØTÜ5ØNÜ5ØNÜ] Where, "" is the shape parameter of gamma distribution, 'a' is the variance and 'g' is the gray level.

Poisson Noise
Poisson Noise is a electronic noise which is a form of ambiguity related with the quantity of the light.This occurs in an image when the limited number of particles that carry energy, such as electrons which is small enough to give rise to measurable variations.Consider a light combination of photons coming out of a source and striking a point which creates a evident spot, the physical process which governs the light emission are such that those photos which are emitted from the light source hits the point many times but to create visible spot billions of photons are needed.however, if the source is not able to emit handful number of photons which hits the point every second then this noise is caused.
The formula of adding the Gaussian Noise to an image is: J = imnoise (I,'poisson') where I is double precision, then input pixel values are interpreted as means of Poisson distributions.

blurred Noise
Blurred Noise is caused due to the light intensity and external factors.Capturing reasonable photos under low light conditions using a handheld camera can be annoying experience.Often the photos taken are blurred or noisy.These kinds of images containing hazy and blurred pixels are referred to as Blurred Noise which is present in the image.

Literature Review
Noise reduction is a very essential step in digital image processing for getting better quality images.Medical imaging is a valuable tool in the field of medicine.Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Ultra Sound imaging (USI) and other imaging techniques provide more effective information about the anatomy of the human body, during the diagnosis process 9 .In the medical field the Surgeons always desire for enhanced medical images for the diagnosis because most of the time the images are not perfect and are deteriorated by many internal and external factors.The low quality of medical images causes difficulty for the Surgeons at the time of diagnosis or interpretation.A quality image is needed by Biometric Identification and Authentication Systems to aim at consistent and exact outcomes so that it can be helpful for universalperson trained in medical science to study the prodrome of the patients.The quality of the MRI and brain imageis obtained by the noise free images to get the better result and increased in accuracy of the result.Many filters are applied to get the best possible result for the noises present in the image like Weiner filter, Median filter etc. Weiner filter and Median filter gives the best result compared to the other filters for the Speckle Noise, Gaussian Noise and Poisson noise as well which are present in an image 10 .Some filters work best for the specific Noises like Salt Pepper, Gaussian, Speckle, Blurred Noise etc. and later in the experiment it will be briefed.The advantage of Median filter is to remove cells which are distant from the observations experimented without reducing the quality of the image and the disadvantage is that while smoothing the noise in the image loses its quality on the edges and boundaries it also erases the details in the image.Gaussian filter advantage is for peak detection and disadvantage is it reduces the details from an image 11 .

RESULTS
The implementation of various de-noising algorithms with different filters has been carried out using MATLAB.here the images considered are MRI brain images, in RGB and gray scale affected by noises like PoissonNoise, Speckle Noise, Gaussian Noise and Salt and Pepper.20 MRI medical images is been taken for the experiment of the noise and its removal.These images have been processed in the MATLAB by adding different noises to an image.After adding the noise to an image different noise filtering algorithm is used to remove the noise from an image.

Speckle Noise
In medical ultrasound imaging speckle noise is inbuilt property which normally tends to reduce the quality of the image, contrast and pixels.Gaussian distribution which is also known as normal distribution whose probability is equal to statistical noise known as Gaussian Noise.Due to poor illumination andgreat extent of temperature or transmission of particles in an electronic image it fails to meet up the requirement for the clear image this noise is caused.Byusing a Weiner filter this noise can be reduced to very much extent.Same filters are used to check out the best filter for this noise.The best filter was Weiner filter but the minimum size value after using the different filters is given by all three filters shown in fig 5.Each filter is used on all the images, their outcome is noted and compared with the other entire filter applied on the same images.Thecircularly symmetric Gaussian behavior is found in the mellow ultra-sound speckle echo for marginal statistics which is similar to the laser speckle for monochromatic illumination 12 .
The result which is achieved showed as follows: Poisson is also known as shot photon noise is the noise which is caused when sensor is not sufficient to provide detectable statistical information even after sensing number of photons 13 .This kind of noise is a type of electronic noise which occurs in an image due to small number of particles that carry energy 14 .This noise was added to the MRI images.

Salt Pepper Noise
Fifth Noise is the Salt and Pepper Noise which is added to all the MRI images and the filtered applied on these images is Weiner Filter,

CONCLUSION
In this work we have taken twenty different medical images like MRI for doing our experiment for noise removal.We have added salt pepper, Gaussian, speckle, blurred and poison noise to the images and also removed these noises from the above medical images by applying the various filtering processes like Median, Gaussian and Weiner Filtering techniques.In order to achieve accurate results for the given application it is mandatory to get good and clear images.The results are examined and compared with ordinary pattern of noises; these are examined through the quality pixels, size, clarity and histogram of these images.From this experiment we come to

Fig. 1 :
Fig. 1: Speckle Filtered Image size given by three filters

Fig. 3 :Fig
Fig. 3: blurred Filtered Image size given by three filters

FigFig. 5 :
Fig.4: (a)blurred MRI image(b) De-blurred Image (a) (b) Poisson distribution generally satisfies in many images which are having Poisson noise and also come across normally distributed and additive noise.Example radiography images and MRI images.Depending on the image intensity the magnitude of Poisson noise varies across an image which makes hard to remove the noise.All three filters Gaussian filter, Weiner filter, Median filter to remove Poisson Noise.The minimum size value after the filtration of the image was a contradiction between Gaussian and Median filter but the best clarity was achieved by Median filter with our experiment.