You are on page 1of 7

Disponible en ligne www.sciencedirect.



Procedia Computer Science 70 (2015) 579 - 585

4 e Confrence internationale sur l'informatique cologique et des systmes de communication, ICECCS


ROI base sur la compression d'images mdicales pour Tlmdecine


Manpreet Kaur une, VikasWasson une

une Dpartement des sciences informatiques, Universit Chandigarh, Gharuan, Mohali140413, Inde


As the medical imaging and telemedicine has been developing on large scale. With the increasing demand of storing and sending the medical images results in
lack of sufficient memory spaces and transmission bandwidth. To fix these issues compression was introduced. Over the past few years in medical imaging
lossless compression schemes are under intensive interest because there is no loss of information. The only small part is more useful out of the whole image.
Region of Interest Based Coding techniques are more considerable in medical field for the sake of efficient compression and transmission. The current work
begins with the pre-processing of medical image. Then segmentation is applied to divide the image into two parts i.e. ROI and non ROI. Finally compression is
performed to reduce the storage and network bandwidth.In this paper Fractal lossy compression for Non ROI image and Context tree weighting lossless for ROI
part of an image have been proposed for the efficient compression and compared with other such as Integer wavelet transform and Scalable RBC. 2015 The
Authors.Published by Elsevier B.V.

2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (
Peer-review under responsibility of organizing committee of the International Conference on Eco-friendly Computing and Communication Systems
(ICECCS 2015).
Peer-review under responsibility of the Organizing Committee of ICECCS 2015

Keywords:Telemedicine ,Compression, Region of Interest,FractalCompression,Context Compression ;

1. Introduction

Telemedicine is a sub-set of eHealth electronic process in wellbeing utilizing data and correspondence advances (ICT) for Teleconsultation,
Medical Second Opinion Service or Tele-expertise, Tele-monitoring, remote taking care of or Tele-assistance.In it, there is transfer or exchange
of medical image data such as X-ray, ultrasound images for the diagnostic purpose 1. The main goal of telemedicine to use the advance
technology to improve the health of the patients in those areas where geographical distance becomes the barrier.Telemedicine for the most part
has two essential capacities. First is the Transfer of Patient's therapeutic information as an option of patient moving starting with one spot then
onto the next 3. Video Conferencing between patient end and master specialists for discussion,

1877-0509 2015 Les auteurs. Published by Elsevier BV Ceci est un article d'accs ouvert sous la licence CC BY-NC-ND (
Examen par les pairs sous la responsabilit du Comit d'organisation de ICECCS 2015 doi: 10.1016 /
580 Manpreet Kaur and Vikas Wasson / Procedia Computer Science 70 ( 2015 ) 579 585

treatment and follow up. After direct communication, at the multispecialty hospital senior doctor checks the information and sends the report
back to the nearby, who gives treatment to the patient. In this information and communication technologies are used .It reduces the stress
level in patients due to the travel time and its expensive. It is more economical as it decreased go time 8.

1.1. Need of Compression

In telemedicine, patients medical information transfer from one multispecialty hospital to the local hospital. Hospital stores this information for
the future use. But the size of medical image is large or can be bmp images. Multispecialty hospital produces large no. of images per patient.
The amount of images produces by the hospital takes the 5 to 15 GB of storage space per day 2. It is too difficult to manage the storage
system in the hospitals because it is mandatory for hospital to store the medical record of each patient. Moreover to send these images over
the network needs high bandwidth 3. It may increase the transmission cost. In rural area, there are many network issues which may cause the
problem in transmission of data. To deal with the above problems; compression was introduced.

Compression lessens the size of images. It is of two types i.e. lossy and lossless. Lossy techniques are used where loss can be accepted and
Lossless techniques are used for the application that cannot afford any loss of information for example in medical field 4.

Lossless compression for medical images has proposed in this work and compared with other techniques i.e. Integer wavelet transform and
Scalable RBC. The proposed algorithm includes two compression techniques i.e. Fractal is used for the non ROI part of an image and Context
tree for ROI part which cannot afford any loss 4. This paper is structured into four sections. First section includes the introductory part and
second section explains about the details about the proposed techniques in this paper. Third section includes the Experimental results.
Conclusion and future scope of proposed work include in the last section of a paper 10.

2. Region of Interest

The medical image includes three parts in image. These are ROI (region of interest), non ROI and background. These part have their own
advantanges.ROI is the most critical part of the image that located over very small regions of the image. Non ROI is also included so that user
can easily find out the most critical part from the whole image 1.
Part other than image contents is known as background and most ignored part of the image. In medical field, these critical parts needed to be
compressed with high quality compression without any loss than other parts of image. The critical parts from the image obliged to be transmitted
first or at higher need amid the transmission for telemedicine purposes. Figure 1 is given below that shows the three different parts of the medical




Fig.1. Different parts of an image

2.1 Related Work

Dr.S.Shenbaga Devi and K.Vidhya 3 in 2007 have described the major issues faced during the transmission of data. They recommend the
wavelet based techniques with SPIHT. SPIHT works more effectively than JPEG as from results. IWT based compression is more useful in
case of DICOM (Digital Imaging and Communication in
Manpreet Kaur and Vikas Wasson / Procedia Computer Science 70 ( 2015 ) 579 585 581

Medicine) images. It gives calculations faster than the traditional DWT (K Vinayak and Ashok M Sapkal 2013). Low computational complexity
because IWT generates only integer numbers. It is reversible with no loss.Peng Wuet al. 7 in2011 have proposed a method. It is based on
carefully choosing the combinations of color space transform and wavelet transform. The paper started with a color space transform and
wavelet transform, and then experiments have been made to select the most outstanding combination of these two transforms. To balance
the quantity of information in the different color component, a weighting matrix is designed which improves objective quality of reconstructed
images. Context-based and regions of interest (ROI) based approach was proposed by M.A. Ansari 18 in 2009.This is especially to compress
vascular type medical images where, there is requirement of high resolution and contrast.

Although in Related work, both lossy and lossless compression methods have been discussed but these techniques have following flaws.

1 These require further coefficients in case of MAXSHIFT, Vector Quantization, EZW-based and other ROI
methods to decode the objects. This may increase the complexity 1.
2 The JPEG 2000 standard sometimes behaves as a lossy ROI compression 3.
3 The previous JPEG algorithm is lossy - when an image is converted to JPEG, some of the information in
the image is lost 4.
4 Discrete Wavelet Transform (DWT) based ROI coding techniques has a drawback that it increases
computational complexity as it works with floating point 4.
5 Compression using Wavelet Transform (WT), results in a lossy scheme 6.

To overcome these flaws new technique has been proposed.In this paper Fractal lossy compression for Non ROI image and Context tree
weighting lossless for ROI part of an image have been proposed for the efficient compression.

The current work begins with the pre-processing of medical image to remove the noise from an image. Then segmentation is applied to divide
the image into two uniform parts i.e. ROI and non ROI 17. Finally compression is performed to reduce the storage and network bandwidth. Here
two compression methods are used. After these results are compared with the IWT and Scalable RBC compression techniques with help of
Compression Ratio, MSE and PSNR.The previous methods encode image by pixel based systems but Fractal compression is based on
image structure. It is lossy in nature so non ROI is compressed with it.In it decoding stage is quick. Fractal Compression works on following

1. In first step selected image is divided into blocks. These blocks are non overlapping. These are knows as parent blocks (PB). The
image processing techniques are used for dividing the whole image based on color separation, edge detection and texture analysis.

2. Further all parent blocks are divided into 4 sub blocks. These blocks known as child blocks (CB).
3. Then each child block is compared with its parent block for finding the closest match of 1 child block out of 4 child blocks.

4. Apply transformation on that child block.

5. The information regarding the location of parent block, child blocks and transform component stored in a file.

6. Repeat this procedure for each child block to get compressed image.
7. During decoding time, use blank starting image of same size as original image and read each stored child block.

8. Apply the stored information of that child block to blank image to get reconstructed image.
9. Repeat this process at decoding end until acceptable image quality is achieved.

Context Tree is prediction algorithm. It gives the both theoretical and practical guarantees for the performance. The CTW technique is a "group
technique," blending the expectations of numerous hidden variable request Markov models, where each such model is built utilizing
zero-request probability estimators.
582 Manpreet Kaur and Vikas Wasson / Procedia Computer Science 70 ( 2015 ) 579 585



ROI Classification

(Fractal) Non-ROI

Lossless Compressi ion Losssy Compression
(Context Tree)


Decompression L
Lossless ROI Image e

Fig.2. Block k Diagram of Propo sed method

3. Exxperimental Re esults

In thi is section imple od ementation of t r the proposed m method is discus ssed .The exper riments results iginalshows the prop posed
metho works bette re 4 than the exis sting techniques f s. The figure 3 ROI shows the ori image be After efore any opera ingation.
Figur shows the c the classification of ROI and non from the o done. original image. 5 performi the compres mage ssion h
with techniques dec ng compression is he Figure 5 tion usingshows the re osed econstructed im with both T+SRBC the
proposed t osed and
propo existin own in the technique. Th re 5.rate of distort propo method is lesser than IWT techni iques
as shofigur

Fig.3.Original Im mage

Fig.4.RROI and non ROI im mages

Manpreet Kaur and Vikas Wasson / Procedia Computer Science 70 ( 2015 ) 579 585 583

Image originale IWT + RBC Fractal + CTW

Fig.5. Reconstructed Image

3.1 Tables

Table1.Comparison of Proposed Method (Parameter CR)

Ensemble de donnes d'image fractale + CTW IWT + GRM chantillon 1

86,4319 81,5708

Echantillon 2 91,9231 87,0620

L'chantillon 3 93,1256 88,2645

L'chantillon 4 90,8100 85,9489

Echantillon 5 85,7119 80,8508

moyenne CR 89,6005 84,7394

Tableau 2. Comparaison de la mthode propose (paramtre MSE)

Ensemble de donnes d'image fractale + CTW IWT + GRM chantillon 1

1,8295 11697

Echantillon 2 0,4567 8316

L'chantillon 3 0,1561 16143

L'chantillon 4 0,7350 16938

Echantillon 5 2,0095 13634

moyenne MSE 1,03736 13345,6

Tableau 3.Comparison de la mthode propose (paramtre PSNR)

Ensemble de donnes d'image fractale + CTW IWT + GRM chantillon 1

49,4178 11,3602

Echantillon 2 55,4447 12,8419

L'chantillon 3 60,1073 9,9613

L'chantillon 4 53,3783 9,7523

Echantillon 5 49,0103 10,6947

Moyenne PSNR 53,27168 10,92208

A partir des valeurs tabulaires les graphiques ont t tablis comme indiqu ci-dessous. Le graphique est trac entre cinq images chantillons et leurs taux de compression
respectifs pour les techniques proposes et existantes.
584 Manpreet Kaur et Vikas Wasson / Procedia Computer Science 70 (2015) 579 - 585

Fig.6. Comparison of Fractal Compression+ CTW with IWT+SRBC (Parameter CR)

Graph shows high compression ratio for Fractal+CTW but without degrading quality of the images. But in case of IWT and SRBC, It decreases
the picture quality due to high compression.

Fig.7. Comparison of Fractal Compression+ CTW with IWT+SRBC (Parameter MSE)

The above graph shows the Mean Square Error in the images high for the IWT+SRBC as compared to the proposed method. From graphs 1,
2 and 3, we can see the values of compression ratio and PSNR are high in case of Fractal+CTW compression method. Below Graph is plotted
to show comparison with PSNR parameter. PSNR value is high when mean square error is less. But in case of IWT+SRBC the value of MSE
is high so low PSNR value. High PSNR value means high quality image.

Fig.8.Comparison of Fractal Compression+ CTW with IWT+SRBC (Parameter PSNR)

Manpreet Kaur and Vikas Wasson / Procedia Computer Science 70 ( 2015 ) 579 585 585

4. Conclusion and Future Scope

Image compression extensively used in real time applications where data is transmitted over network. In this paper, Context Tree
lossless compression of ROI and Fractal compression for non-ROI area is implemented. From the outcomes, the proposed strategies have
demonstrated a low Mean Squared Error rate, high PSNR and High Compression Ratio when contrasted with the past methods i.e. IWT and
Scalable RBC. According to the experiments performed so far as, it proved that the proposed techniques Fractal and Context tree provide
more accurate and faster results than the previous techniques i.e. IWT and Scalable RBC. The future work is to increase PSNR, CR values
and work with more parameters.


Author would like to thank the Chandigarh University Gharuan, Mohali, India for supporting and providing help to this research. Author
would also like to thank all the reviewers for their valuable comments and constructive suggestions.


1. BairagiVinayakK., and Ashok Sapkal M. ROI-based DICOM image compression for telemedicine. Sadhana, 2013; Vol. No.38, pp.123-
2. Brindha B.&Raghuraman G. Region based lossless compression for digital images in telemedicine application. International Conference inCommunications and Signal
Processing (ICCSP), 2013; pp. 537-540.
3. Devi S. S. &VidhyaK.. Development of medical image compression techniques. In IEEEConference on Computational Intelligence and Multimedia Applications, 2007; Vol. 3,
pp. 97-101.
4. P. Wu P., Xie K., Yu H., Zheng Y. and Mao Yu W.A New Preprocessing Algorithm Used in Color Image Compression.. Advances in Future Computer and Control Systems. Springer
Berlin Heidelberg, 2012;pp.465-471.
5. Ueno I., & Pearlman W. A. Region-of-interest coding in volumetric images with shape-adaptive wavelet transform. In Electronic Imaging,
International Society for Optics and Photonics in 2003; pp. 1048-1055.
6. Sanchez, V., Abugharbieh R.&Nasiopoulos P. 3-D scalable medical image compression with optimized volume of interest coding. Medical Imaging, IEEE Transactions, 2010;
Vol. No.29, pp.1808-1820.
7. Mohandass D. and Janet J. An improved three pattern Huffman compression algorithm for medical images in telemedicine. Information
Processing and Management. Springer Berlin Heidelberg, 2010: pp.263-268.
8. Arthur, A., & Saravanan, V. technique efficace de compression d'images mdicales pour la tlmdecine application en ligne et compte tenu de hors-ligne.
Confrence internationale onComputing, Communication et Applications (ICCCA), 2012: pp. 1-5.
9. K. Krishnan, Marcellin MW, Bilgin A., et Nadar MS transmission efficace des donnes compresses pour la visualisation de volume distance.
Imagerie mdicale, IEEE Transactions on, 2006; Vol. No. 25, pp.1189-1199.
10. M Kaur et Wasson V, Rgion d'intrt sur la base des techniques de compression pour Tlmdecine application, Journal international de recherche
en gnie lectronique et informatique , 2015: Vol. 3, pp. 63-67.
11. V. Sanchez et R. Abugharbieh symtrie base volutive Lossless Compression de donnes d'images mdicales 3D, IEEE Transactions on
L'imagerie mdicale, 2009; Vol.28, n 7, pp. 1062-1072.
12. Gonzalez Rafael C. Digital image processing. Pearson Education India, 2009.
13. Schelkens P., Munteanu A., Barbarien J., Galca M. , Giro-Nieto X., &Cornelis, J. (2003). Wavelet coding of volumetric medical datasets.
Medical Imaging, IEEE Transactions on, 2003; Vol. No. 22, pp. 441-458.
14. Kamsu-Foguem B., and Foguem C. Telemedicine and mobile health with integrative medicine in developing countries. Health Policy and
Technology 3, 2014; Vol.No. 4, pp.264-271.
15. RazaviHesabi Z., Sardari M., Beirami A.,Fekri F.,Deriche M. & Navarro A. A memory-assisted lossless compression algorithm for medical images. IEEE International Conference In
Acoustics, Speech and Signal Processing (ICASSP), 2014; pp. 2030-2034.
16. Miaou S. G., F. S. Ke& S.C Chen, A lossless compression method for medical image sequences using JPEG-LS and interframe coding.
IEEE Transactions on Information Technology in Biomedicine, 2009; Vol. No. 13, pp. 818-821.
17. Hsu, Wei-Yen. Segmentation-based compression: New frontiers of telemedicine in telecommunication. Telematics and Informatics, 2015;
Vol.No.32, pp.475-485.
18. Ansari M. A. and Anand R. S. Context based medical image compression for ultrasound images with contextual set partitioning in hierarchical trees algorithm. Advances in
Engineering Software 2009; Vol.No. 40, pp.487-496.
19. Shah R., & Sharma P. Performance Analysis of Region of Interest Based Compression Method for Medical Images. 2014 Fourth
International Conference in Advanced Computing & Communication Technologies (ACCT) IEEE, 2014; pp. 53-58.
20. M. Firoozbakht, J. Dehmeshki, Martini M., Ebrahimdoost Y., Amin H., Dehkordi, M. & Qanadli, compression SD des images mdicales numriques bass sur plusieurs rgions d'intrt. Quatrime
Confrence internationale dans Socit numrique, (ICDS'10), 2010; pp. 260-263.