Two Dimensional Mri Image Analysis by Using Oversampled Fir Filter Bank for Perfect Reconstruction

In this paper we propose oversampled three channels FIR filter banks [FB]. Channels are selected likewise which gives the result near perfect reconstruction filter bank. The inband aliasing is significantly reduced by selecting proper frequency spectrum and design of filters of filter banks. This design gives the result of oversampled FIR filter bank for two dimensional MRI image with peak signal to noise ratio and histogram of images. By applying wavelet transform energy is preserved at output of oversampled filter banks. By using local thresholding segmentation, major components of images are exposed, which detect abnormality i.e. occurrence of cancer tissues in image. I] INTRODUCTION: Biomedical images are of significant interest because they can be a useful tool when diagnosing and analyzing functions and diseases of the brain, as well as other body parts can be examined for example the kidneys. So how is it that the body can be analysed by magnetic resonance? Well, the biological tissue contains a lot of hydrogen atoms which are possible to detect. Nuclear magnetic resonance is a technique in which a electromagnetic .field is applied to the sample, in this case the brain. The nuclei of the hydrogen atoms in the biological tissue align themselves to the magnetic field, after this a radiomagnetic pulse will raise their energy level further, when the pulse ends they will relax and during the relaxation this energy will be transmitted from the atoms. The transmitted signal will be detected by the equipment and processed further into the pixels that make up the biomedical image. This paper is more focused on MR images, but in fact the methods described here can be applied to any kind of image. During recent years ,the efficiency of image coding algorithm is improved significantly. Typically signal decomposition is performed by using discrete FIR filter bank. Uniform FIR filter bank have variety of applications in speech processing, image processing, and signal processing. Applications of oversampled and nonuniform filter banks can be found in those area of signal processing where one is interested in making modification in signal processing to signals in certain frequency bands. Recently perfect reconstruction condition for oversampled and nonuniform filter bank has been derived. Because of real valued subband signals, these filter banks are more suitable for spectral modification.[1]. Basic goal of this paper is Image de-noising and decomposition using selected filtering methods. Good algorithms for de-noising are of the essence when it comes to handling MR images. FIR filters are used to decompose and de-noise the image using filter bank. The design of three channel oversampled FIR two dimensional filter bank is implemented, which satisfies all the properties of perfect reconstruction of filter bank. This filter bank TWO DIMENSIONAL MRI IMAGE ANALYSIS BY USING OVERSAMPLED FIR FILTER BANK S. R. Chougule, et al. 106 passes all the frequency components without any loss and as well as to filter all noise with each channel and finally de-noised image without inband aliasing produced at the output of filter bank. To detect cancer tissue in the 2D MRI image, denoised image is applied for image segmentation using thresholding of image. For image segmentation we have been appling thresholding i.e.local thresholding and global thresholding. Basic goal of this paper is to design proposed three channel oversampled filter bank which is applied for two dimensional MRI image to filter out all the noise and inband aliasing in channels and give output near perfect reconstruction of filter bank. The de-noised image which is obtained from FIR filter bank applied for segmentation and thresholding which are able to enhance feature of MRI image. The aim of our study was to assess the accuracy of thin-section MRI performed with a phased-array coil as a technique for the preoperative evaluation of pelvic anatomy and tumor extent in patients with rectal cancer[2]. With the discussion with radiologist comparison of normal MRI image and MRI with tumor images are selected for analysis. This technique is suitable for any 2-D image analysis. In section II filter bank description is explained. Section III shows design procedure for FIR filter bank . Section IV gives details about image segmentation. II] Oversampled three channel Filter Bank: X(k) Y(k) Figure.1: Three Channel filter bank The main problem of subband adaptive filtering is the”inband” alias which occure if a real valued analysis filters Hi(z), whose channel is subsampled by Si, contains normalized frequency points 2лl/Si, l=1,....Si – 1 Ref [3]. Therefore to avoid inband alias, all the analysis filters need to have spectral nulls at these frequencies, if the same subsampling ratio Si is to used for all channels. So filterbank proposed is to remove inband aliasing by choosing different subsampling ratio Si for different channel[4] . We proposed three channel oversampled filter bank with same subsampling ratio as shown in Figure [1].Here every analysis filter has to be placed in the frequency domain such that the resulting signal does not violate the sampling when subsampled by a factor Si and that the analysis filters have to be placed such that all frequencies are covered by at least once to filter in order to allow reconstruction. This filter bank is simplest possible filter bank that uses different subsampling ratios. This filter bank preserves the property of alias free output of two dimensional image . In this design H0(z) is lowpass filter and H2(z) is highpass filter which covers all frequency components of input signal except frequency pi*1/2 i.e.0.5 normalised frequency which is shown in Fig.1[1]. H1(z) is selected as a band pass filter of passband normalized frequency range shown in Fig2. H0(z), H1(z) and H2(z) are downsampled by 2. All frequencies must be covered by at least one filter. To fill the spectral gap between H0(z) and H2(z), one bandpass filter H1(z) is selected. Frequency spectrum of above figure is shown in Fig.2. Design of this filer bank using 1D direct formII transposed structure FIR filter design technique which is transformed into two dimensional FIR filter using frequency transformation technique. . H0(z)