23-07-2012, 03:47 PM
A Modified Method for Speckle Noise Removal in Ultrasound Medical Images
A Modified Method for Speckle Noise Removal in Ultrasound Medical Images.pdf (Size: 273.73 KB / Downloads: 60)
Abstract
Ultrasound images contain speckle noise which degrades the quality of the images. Eliminating such noise is an important preprocessing task. This paper describes and analyses an algorithm for cleaning speckle noise in ultrasound medical images. Mathematical Morphological operations are used in this algorithm. This algorithm is based on Morphological Image Cleaning algorithm (MIC) designed by Richard Alan Peters II. The algorithm uses a different technique for reconstructing the features that are lost while removing the noise. For morphological operations it also uses arbitrary structuring elements suitable for the ultrasound images which have speckle noise.
Index Terms—bottom hat, morphology, reconstruction, speckle, structuring element, top hat
I. INTRODUCTION
Ultrasound imaging is widely used in the field of medicine. It is used for imaging soft tissues in organs like liver, kidney, spleen, uterus, heart, brain etc. The speed, low cost of imaging and the portability of scanning machine makes it very popular. The common problem in Ultrasound image is speckle noise which is caused by the imaging technique used that may be based on coherent waves such as acoustic to laser imaging [8][9]. This paper describes and analyses an algorithm for reducing such speckle noise. This algorithm is based on mathematical morphology. It is a modified form of MIC and it is called as MMIC. It differs from MIC by not using the histogram for calculating the threshold of the image. It is also using a different technique for reconstructing the features that are of speckle’s size. Moreover it uses structuring elements which are having arbitrary structures which resemble the shapes of the speckles. This algorithm produces better result when compared to the original MIC in time complexity as well as output quality. This paper is organized as follows. Section II discusses the previous works in the literature. Section III describes the MIC and the modifications done to that in this paper. Section IV gives the modified version of MIC. The results and discussion are given in Section V.
II. RELATED WORKS
Various techniques for speckle noise removal are available
T.Ratha Jeyalakshmi is a Research scholar at Mother Teresa University, Kodaikanal, Tamil Nadu, India. (radha_jeyalakshmi[at]yahoo.com) Dr.K.Ramar is Professor and Head in Department of Computer Science and Engineering, National Engineering College, Kovilpatti, Tamil Nadu, India. (kramar_nec[at]rediffmail.com .
in the literature [11][13][14][17][19][25]. Linear filtering techniques like spatial averaging have blurring effect [4]. Adaptive filtering techniques based on local statistics [2],[9][20] or spectral coefficients [11] are good in preserving object boundaries and small features with speckle size. Application of non linear filters is also available in the literature. Non linear filters based on Mathematical morphology are size and shape sensitive and they are found to be good in removing speckle patterns [4][5][6][7]. Morphological filters use mathematical morphological operations such as opening, closing, top hat, bottom hat etc. [10][12][13][16]. A morphological algorithm known as Morphological Image Cleaning algorithm is good in reducing noise in different types of images including scanner images[3]. This algorithm finds the residual image which is the difference between the original image and the smoothed image. It separates the features from the residual image and puts it back into the original image so that features are preserved. This algorithm is an iterative procedure which works with disk shaped structuring elements with different radius. First it filters the original image repeatedly using Opening Closing and Closing opening (OCCO) filter using disk shaped structuring elements of different radius. The output of each iteration is processed as follows. The positive elements of the residue are put in an array and the other elements are put in a different array. For each array threshold is calculated using the second order moment of the gray level distribution shown by the histogram. Then each thresholded image is cleaned as follows. Rank order filter is applied and the isolated pixels are deleted repetitively until no more isolated pixels are found. Thresholding ends in trimming of bases of features. To recover those bases of the features MIC expands the nonzero regions by skeletonizing and dilating. The first cleaned thresholded image is added to and the second cleaned thresholded image is subtracted from the output of OCCO before continuing the next iteration which takes this result as the input.
III. MODIFICATIONS MADE TO MIC
Since the cleaning process takes much time the algorithm is modified such that it does less processing to get the result. First for filtering using OCCO the arbitrary structuring elements that resemble the shape of the speckle are used. A speckle does not have a regular shape [15]. It is not appropriate to use predefined structuring elements like disk, rectangle, hexagon etc., for morphological processing [9] [19]. Therefore an arbitrary structuring element which resembles the speckle shape is designed. For this random
A Modified Method for Speckle Noise Removal in Ultrasound Medical Images
T.Ratha Jeyalakshmi and K.Ramar
International Journal of Computer and Electrical Engineering, Vol. 2, No. 1, February, 2010
1793-8163
55
speckle samples are taken from different ultrasound images
and by thresholding three structuring elements were
designed as given in fig.1. In MIC for thresholding the image,
histogram of the image is used. In MMIC instead of that the
standard deviation of the pixels in the image is used and from
this the threshold is found. The morphological operations
like opening and closing which are used for filtering in
OCCO will lose the features which are less than or equal to
the size of the structuring element. For reconstructing the
features that are lost while cleaning, opening by
reconstruction and closing by reconstruction are used [1][10].
These are efficient techniques for getting back the lost image
features.