of each stage, all coefficients that have been found to be significant will have their most significant bits recorded.
International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 1
ISSN 2229-5518
Medical Image Compression Using
Multiwavelets for Telemedicine Applications
R.Sumalatha, M.V.Subramanyam
Abstract— In this paper we propose an efficient region of interest (ROI) coding technique based on multiwavelet transform, set partitioning in hierarchial (SPIHT) algorithm of medical images. This new method reduces the importance of background coefficients in the ROI code block without compromising algorithm complexity. By using this coding method the compressed bit stream are all embedded and suited for progressive transmission. Extensive experimental results show that the proposed algorithm gives better quality if images using multiwavelets compared to that of the scalar wavelets. The performance of the system has been evaluated based on bits per pixel (bpp) , peak signal to noise ratio (PSNR)and mean square error (MSE).
Index Terms— Medical image compression, Multiwavelet, Region of interest (ROI), SPIHT, Telemedicine, Scalar wavelets, PSNR, MSE.
—————————— ——————————
NALYSIS and compression of medical imagery is an important area of biomedical engineering. Medical image analysis and data compression are rapidly evolving field with growing applications in the healthcare services e.g teleradiology, teleconsultation, e-health, tele- medicine and statistical medical data analysis[1]. For the telemedicine, medical image compression and analysis may even be more useful and can play an important role for the diagnosis of more sophisticated and complicated images through consultation of experts[2]. In medical image compression diagnosis and analysis are effective only when compression techniques preserve all the rela- vant and important image information needed. This is the case of lossless compression. On the otherhand lossy compression is more efficient interms of storage and transmission needs but there is no gauranty to preserve the characteristics needed in medical diagnosis [3]. To avoid the above problem, there may be third option that the diagnostically important region (ROI) of the image is lossless compressed. ROI, a segmentation approach can be used to extract the ROI. These regions are very useful for diagnosis.Hence, the ROI must be compressed by a
lossless or a near lossless compression algorithm.
Wavelet based techniques are latest development in
the field of medical image compression. In most cases, the
wavelet transform produces floating point coefficients
and although this allows perfect reconstruction of the
original image in theory, the use of finite precision arith-
metic together with quantization results in a lossy scheme.Another way to achieve improved compression
————————————————
results over wavelets is to use integer wavelets. Intger wavelet transform demonstrate a significant improve- ment in reconstructed image quality over the wavelet transform for medical images[4]. For best performance in image compression, wavelet transforms require filters that combine a number of desirable properties, such as orthgonality and symmetry. However, the design possi- bilities for wavelets are limited because they cannot si- multaneously possess all of the desirable properties. The relatively new field of multiwavelets shows promise in obviating some of the limitations of wavelts. Multiwave- lets offer more design options and are able to combine several desirable transform features.
This paper is organized as follows: section two re- gion of interest section three describes wavelet and integer wavelet transforms, section four describes mul- tiwavelet, section five describes SPIHT algorithm, sec- tion six experimental results section seven describes conclusion and future work.
ROI coding is one of the most important features provided by JPEG-2000. It allows, imposing heteroge- neous fidelity constraints to different regions of the image rather than encoding the entire image as a single entity. This property is especially useful for image cod- ing applications, where the image consists of regions that can be encoded at different bit rates, such as com- pression of medical images [5]. For most medical im- ages, the diagnostically significant information is loca- lized over relatively small regions of interest. In this case, region-based coding for better utilization of the available bit rate since the high quality should be maintained only for the aforementioned diagnostically significant regions and the rest of the image can be en- coded at a lower bit rate. Once the region of interest is selected efficiently, the significant region is trans-
IJSER © 2011
International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 2
ISSN 2229-5518
formed using lossless integer multiwavelet transform filter. Then the transformed images are encoded using SPIHT algorithm.
Wavelet based techniques are the latest development in the field of image compression. It offers multiresolu- tion capability that is not available in any of the other methods. The low frequency components in the signal that are spread out in time, as well as the high fre- quency components that are localized in time are cap- tured by a variable-length window[6]. The window is shifted by different units of time in a discrete manner, thus covering the entire signal. The wavelet transform (WT), in general, produces floating point coefficients. Although these coefficients can be used to reconstruct an original image perfectly in theory, the use of finite precision arithmetic and quantization results in a lossy scheme. Recently reversible integer wavelet transforms have been introduced [7]. Lifting is a technique used in constructing second generation wavelets, entirely in the spatial domain. The first generation wavelets are translates and dilates of a single mother wavelet, and Fourier techniques are essential in their construction. The lifting scheme does not use the Fourier technique. It is a very fast method compared to the first genera- tion wavelets. Moreover, the inverse transform is ob- tained by replacing addition with subtraction and re- versing the operations in the forward transform.
Multiwavelets are defined using several wavelets with several scaling functions. Multiwavelets have several ad- vantages in comparision with scalar wavelet. The features such as compact support, orthogonality, symmetry, and higher order approximation are known to be mportant in signal processing. Multiwavelets provide one alterna- tive to the wavelet transform. Multiwavelets are very similar to wavelets but have some important differenc- es. In particular, where as wavelets but have an asso- ciated scaling function and wavelet function, multi- wavelets have two or more scaling and wavelet func- tion. Multifilter construction methods are already be- ing developed to exploit the useful properties such as ortogonality, symmetry and high order of approxima- tion [8]. However, the multi-channel nature of muti- wavelets also means that the sub band structure result- ing from passing a signal through a multifilter bank is different. A single level of standard wavelet decompo- sition splits the input signal into lowpass and highpass coefficients through filtering and down sampling. A multi-level wavelet filter bank involves iterating the low pass - high pass filtering and down sampling pro- cedure only on the output of the lowpass branch of the previous stage. During a single level of decomposition
using a scalar wavelet transform, the 2D image data is replaced with four blocks corresponding to the sub bands. The multiwavelets used here have two chan- nels, so there will be two sets of scaling coefficients and two sets of wavelet coefficients since multiple ite- rations over the lowpass data are desired, the scaling coefficients for the two channels are stored together. Likewise, the wavelet coefficients for the two channels are also stored together.
The main idea of multi filter bank is same as filter bank of scalar wavlets.A time scale representation of digital signal is obtained by using different filtering techniques.In case of scalar wavelets the signal is passed through the high- pass and lowpass filters to analyze the high and low fre- quency components.The resolution of signal, which is measure of detail information in the signal, is changed by the filtering operations and the scale is changed by up- sampling and downsamplingoperations.On the other hand a multi filter bank is a combination N ordinary fil- ters,each operating on the separate data stream.
Wavelt produces one lowpass subband and one highpass subband in each dimension. Since multiwavelets produc- es two lowpass subband and two highpass subband in each dimension.The multiwavelets used here have two channels, so ther will be two sets of scaling coefficients and two sets of wavelet coefficients.During a single level of decomposition using a scalar wavelet transform, 2D image data is replaced by four blocks corresponding to the subbands representing either lowpass or highpass filtering in each direction.
The set partition in hierarchical trees (SPIHT) cod- ing algorithm was proposed by said and Pearlman [9]. SPIHT quantizer achieves good performance by ex- ploiting the spatial dependencies of pixels in different sub bands of a scalar wavelet transform. The method deserves good image quality, high PSNR, Especially for color images. It has been noted [10] that there exists a spatial dependence between pixels in different sub bands in the form of a child-parent relationship. The coder starts with a threshold value that is the largest integer power of two that does not exceed the largest pixel value. Pixel values are evaluated in turn to see if they are larger than the threshold; if not, these pixels are considered insignificant. If a parent and all of its descendents are insignificant, then the coder merely records the parent‘s coordinates. Since the children‘s coordinates can be inferred from those of the parent, those coordinates are not recorded, resulting in a po- tentially great savings in the output bit stream. The parent- child relationship in this scheme shown in fig.1. After locating and recording all the significant pixels for the given threshold, the threshold is reduced
IJSER © 2011
International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 3
ISSN 2229-5518
of each stage, all coefficients that have been found to be significant will have their most significant bits recorded.
Fig.1.Illustration of parent-child relationship
The original MRI image is taken as test image.Fig.2. (a), (b), and(c) shows that the original image, mask, and background suprresed images.The multiwavelet filters are used in this work are ‗GHM‘,‘SA4‘,‘CL‘ and also used
‗HAAR‘ scalar wavelet.Table 1 gives the result of compa- rision of different multiwavelet and scalar wavelet at dif- ferent bpp interms of PSNR.Table 2 gives the result of comparision of different multiwavelet and scalar wavelet at different bpp interms of MSE.Fig.3 shows the compari- sion of PSNR values using sacalar wavelet and multi- wavelet at different bpp.Fig.4 shows the comparision of MSE values using scalar and multiwavelet at different bpp.
TABLE 1
COMPARISION OF PSNR VALUES FOR WAVELET AND MULTIWAVE- LET AT DIFFERETN BPP
TABLE 2
COMPARISION OF MSE VALUES FOR WAVELET AND MULTIWA-
VELETS AT DIFFERENT BPP
Fig.2 (a) Original image (b) Mask (c) Background suppresed image
Fig.3.Comparision of PSNR values using scalar wavelets and multiwavelets at level 3
IJSER © 2011
International Journal of Scientific & Engineering Research Volume 2, Issue 9, September-2011 4
ISSN 2229-5518
Fig.4. Comparision of MSE values using scalar wave- lets and multiwavelets at level 3
cations, July-December 2009, vol.3, No.2, page no: 237-240.
[5] Ahmed Abu-Hajar and Ravi Shankar, ―Region of Interest Cod- ing using Partial –SPIHT‖, IEEE 2004.
[6] C.S.Gargour and V.Ramachandran, A scheme for teaching
Wavelets at the Introductory level, ASEE/IEEE Frontiers in Edu- cation Conference, 1997.
[7] Calderbank AR, Daubechies I, Sweldens W, and Leo BL: Wave- let Transforms that Map Integers to Integers. Appl Comput Har- mon Anal 1998:5:332-369.
[8] Michael B.Martin and Amy E.Bell, ―New Image Compression
Techniques using Multiwavelets and Multiwavelet packets‖,
IEEE Transactions on Image Processing, vol.10, No.4, April
2001, pp.500-511.
[9] Said and W.A.Pearlman, ―A new, fast, and efficient image co- dec based on set partitioning in hierarchical trees,‖ IEEE Trans.Circuits Syst.Video Technol., vol.6,pp.243-250,June 1996.
[10] J.M.Shapiro, ―Embedded image coding using zerotrees of wavelet coefficients,‖ IEEE Trans. Image Processing, vol.41,
pp.3445-3462, Dec.1993.
Fig.5 (a) Original image (b) Reconstructed image using
SA4 multiwavelet at level 3
The performance of the multiwavelets in general depends on image characteristics.Multiwavelets is superior to cap- ture the edges than wavelets.SA4 multiwavelet gives good PSNR at low bitrates.Futurework includes multi- wavelets coefficients are implemented via lifting scheme i.e integer multiwavelet transform.
[1] Stephen Wong et al, ―Radiologic Image Compression-A Re-
view‖, IEEE Trans. on Medical Imaging, vol..83., February 1995. [2] Ali M.,‖Medical Image Compression using SPIHT for Teleme- dicine Applications”, .IEE seminar on Time-scale and Time-
Frequency Analysis and Applications, pp:22/1-22/5,29 Feb.2000.
[3] Liang S., Rangaraj M. Rangayyan, ―A segmentation – based Lossless Image Coding method for high Resolution Medical Image Compression‖,IEEE Trans.on Medical Imaging, vol.16,No.3,1997.
[4] R.Sumalatha, M.V.Subramanyam ―Losssless Compression of
MRI Images using Integer Wavelet and SPIHT algorithm”, In- ternational Journol of Computational Intelligence Research & Appli-
IJSER © 2011