Sunteți pe pagina 1din 4

International conference on Communication and Signal Processing, April 3-5, 2013, India

Region Based Lossless Compression for Digital Images in Telemedicine Application

B. Brindha, G. Raghuraman

AbstractRegion Based Coding Technique is significant for medical image compression and transmission. In medical images, only a small portion of the image might be diagnostically useful. In these regions, lossless compression can helps to achieve high efficiency performance in telemedicine applications. This paper proposes a very efficient and low complexity compression method for Digital Imaging and Communications in Medicine (DICOM) images. Main advantages of Region based coding technique is exploited in this paper. Segmentation of these regions leads to optimal performance. Here the ROI part of the image is identified by manually and combined with effect of Integer Wavelet Transform (IWT). IWT compression technique is useful to reconstruct the original image, reversibly with desired quality. The overall compression process helps to reach a satisfactory level for image transmission in limited bandwidth over a telemedicine application. Index Terms Compression, Discrete Cosine Transform (DCT), Integer Wavelet, Region of Interest (ROI) .



Medical imaging has a great impact on the diagnosis of diseases and surgical planning. However, imaging devices continue to generate more data per patient, often 1000 images or ~500 MB. These data need long-term storage and efficient transmission. Many compression algorithms produce high compression rates in medical images. But the quality of loss is affordable. In telemedicine application field medicine cannot afford any deficiency in diagnostically important regions (’Region of Interest, ROI). An approach that brings a high compression rate with good quality in the ROI is necessary. The general theme is to preserve quality in diagnostically critical regions, while allowing lossily encoding the other regions. The main reason for preserving regions other than ROI is make user as more easily to locate the position of critical regions in the original image and also used to evaluate the interactions if possible with other organs. The evaluation of digital images in the field of telemedicine application leads to apply compression algorithm to medical images. The initial emphasis was on information preserving methods. In means

B.Brindha is pursuing post graduate with the Computer Science and Engineering Department, SSN College of Engineering, Chennai. (e-mail:

G.Raghuraman is with the Computer Science and Engineering Department, SSN College of Engineering, Chennai. (e-mail:

978-1-4673-4866-9/13/$31.00 ©2013 IEEE


of scan pixel difference was researched by Takaya et al in [1].

Assche et al exploit the inter-frame redundancy in [2]. Linear

predictive coding schemes were investigated in [3]. The lossless compression algorithms are resulted in low compression ratios. For better result transform coding schemes such as Principal Component Analysis (PCA) and Discrete Cosine Transform (DCT) are used. To achieve high compression ratio without degrading of quality, ROI method were introduced. Data set is a collection of 2D images, henceforth called slices. The most important drawback of 3D based approaches

in ROI based compression is twofold. First, the image quality

along the three principal 3D axes is not uniform, i.e. the

resolution between the slices is much less than the resolution within each slice. Second, the ROI does not necessarily lie in

a 3D primitive shape such as a cube. As a consequence, a

primitive 3D ROI would occupy a big portion of the data, thereby deviating from our initial objective of high compression rate. To address these problems, a 2D based ROI scheme is explored in this paper. In this paper we propose a new lossless compression method in the image with embedded encoding. The Wavelet Transform (WT), in general, produces floating point coefficients. These coefficients are used to reconstruct an original image perfectly in theory, the use of finite precision arithmetic and quantization results in a lossy scheme. Recently, Integer WT has been introduced to overcome this disadvantage. In this paper we have proposed a method for lossless compression in medical images. The proposed algorithm is denoted as Integer Wavelet Transform (IWT). The paper is organized as follows. Section II gives a detailed description of our proposed method. Section III provides experimental results and section IV gives our conclusion and discussion of summary.


ROI concept is introduced owing to limitations of lossless and lossy compression techniques. Most of the lossless compression techniques, the compression ratio are near to 80% of original size, whereas for lossy encoders the compression ratio is much higher (up to 5-30 %) [1][3] but there may be significant loss in data. This loss may hamper effective treatment, losing diagnostically important parts of the medical image. Hence, there is a need for lossless compression technique which will preserve diagnostically

parts of the medical image. Hence, there is a need for lossless compression technique which will

important part (ROI) as well as provide high compression ratio [1][6]. The functionality of ROI is important in medical applications where certain parts of the image are of higher diagnostic importance than others. For most medical images, the diagnostically significant information is located over relatively small regions, near about 5-10 % of the total area of the image. In such cases, these regions need to be encoded at

a higher quality than the background, lesser important regions. During image transmission for telemedicine

purposes, these regions are required to be transmitted first or


a higher priority. The comparison of different ROI-based coding techniques


summarized [7]. The JPEG 2000 imaging standard has

been tested in previously published works on medical images [15]. A major drawback is that it does not support lossy to lossless ROI compression. An overview of ROI coding scalable techniques applied on medical images is summarized in [6][7], which further proposes an ROI-based coding technique based on Discrete Wavelet Transform (DWT). The observed drawback of this method is that every time, in each

sub-band the information about the ROI shape has to be maintained. This may increase computational complexity. Here in this proposed work, an attempt has been made to reduce this complexity.

A. Image Segmentation

Image segmentation is the process of partitioning a digital image into multiple segments. The main goal of segmentation

is to simplify and/or change the representation of an image

into some meaningful part for analyze the image. It also used

to locate objects and boundaries in images. Simply, image

segmentation is assigning a label to every pixel in an image

such that pixels with the same label share certain visual characteristics. The simplest method of image segmentation is called the thresholding method. This method is based on a threshold value to turn a gray-scale image into a binary image. The background of the given input image is detected by applying suitable threshold values. The background part of the image value is less than the thresholding value.


Adaptive Threshold:

For adaptive thresholding algorithm input is grayscale or color image. After the implementation of algorithm binary image is output of adaptive thresholding. In the given image threshold value for each image calculated. If the pixel value is

below the threshold it is set to the background value, otherwise it assumes the foreground value. In Adaptive thresholding, there are two main approaches to finding the threshold such as, the Chow and Kaneko approach local thresholding. The assumption behind both methods is that smaller regions

in an image are more likely to have approximately uniform



Chow and Kaneko divide an image into an array of overlapping sub-images and then find the optimum threshold for each sub-image. The threshold for each single pixel is found by interpolating the results of the sub-image. The main drawback is that it is computational expensive and not appropriate for real time applications. Local thresholding method used to calculate the intensity values of the local neighborhood of each pixel. The statistic which is most appropriate depends largely on the input image. The mean of the local intensity distribution is T=mean the median value, T=median or the mean of the minimum and maximum values, T = (max + min) / 2

mean of the minimum and maximum values, T = (max + min) / 2 Fig. 1.

Fig. 1. Block diagram of the proposed method


Region Growing Segmentation:

Region growing [9] is a simple region-based image segmentation method. It is also named as a pixel based image segmentation method. In region growing segmentation examines neighboring pixels of initial “seed points” and determines whether that the pixel neighbors added to the region or not. This process is similar to the data clustering algorithm. The first step in region growing is to select a set of seed. Seed point selection is based on some user criteria. The initial region begins as the exact location of these seeds. The region growing segmentation starts from these seed points to adjacent points depending on a region membership criterion such as

Pixel intensity Gray level texture

Color. Since the regions are grown on the basis of the criterion the image information itself is very important. For example, if the

criterion were a pixel intensity threshold value, knowledge of the histogram is useful.

B. Classification

All digital images that contains three parts:


ROI (the diagnostically important part), Non-ROI image part, Background (part other than image contents) ROI Detection:

ROI (the diagnostically important part), Non-ROI image part, Background (part other than image contents) ROI Detection:
ROI (the diagnostically important part), Non-ROI image part, Background (part other than image contents) ROI Detection:
ROI (the diagnostically important part), Non-ROI image part, Background (part other than image contents) ROI Detection:

Initially,this background is made zero using segmentation algorithm.After that the ROI part of the image classified from

the images. For the compression the ROI is detected by an automated process. For ROI detection, an ROI-mask is generated in such a way that the foreground part of the image totally included in this mask [8]. Morphological operations are effectively used to generate an ROI-mask. That means the foreground part of the image contains the value of ’1’ and the background part of the image contains the value of ’0’. Then the mask is logically ANDed with the image to separate the ROI part and non-ROI part.

C. Integer Wavelet Transform

Integer wavelet transform (IWT) [11] is lossless compression [13] for medical images. Traditional WT, generally produces the floating point coefficients for the given signals or pixels. These coefficients are useful in reconstruction of original image in lossy scheme [14]. Recently, reversible integer WT introduced for lossless compression. Lifting provides an efficient way to implement the WT and the computational cost of lifting is 30% less than traditional WT. In Integer wavelet transform, reversible IWT [10][12], being simple. The advantages of IWT are:

It is faster than other traditional WT. There is no need of temporary memory.

It generates integer coefficients. So it has low computational complexity as compared with other WT. It is completely reversible.

D. Algorithm

A new algorithm for implementation is presented as, Read the image from and get dimensions for given input image. Apply segmentation algorithm and separate background from image. Detect ROI part of the image. Separate ROI and Non-ROI of the image. Apply compression algorithm.


The data set containing 300 images used for this experiment. The data set include the X-ray images of spinal cord, lungs, skull, etc. The size of the images is around 512X 512 pixels. The background parts of the input images are separated by applying threshold segmentation algorithm. The ROI part of

the image separated from image after that the compression algorithm that can be applied on that ROI part of the image. Then the reverse part of the compression algorithm is used to obtain original image.

the compression algorithm is used to obtain original image . Fig. 2. Original Image Fig. 3.

Fig. 2. Original Image

Fig. 3. Adaptive Threshold

Adaptive thresholding algorithm identifies the background of the images using threshold values. Fig. 3 shows the result obtained after applying adaptive thresholding algorithm.

obtained after applying adaptive thresholding algorithm. Fig. 4. Original Image Fig. 5. Region Growing Fig. 6.

Fig. 4. Original Image

Fig. 5. Region Growing

algorithm. Fig. 4. Original Image Fig. 5. Region Growing Fig. 6. ROI Detection Region thresholding algorithm

Fig. 6. ROI Detection

Region thresholding algorithm identifies the background of the images using some initial seed point values. Fig. 5 shows the result obtained after applying region growing thresholding algorithm. Here ROI of the image identified by manually. Fig. 6 shows the result selection of ROI part. The DCT is used to convert a signal into integer coefficients. The reverse process, that of transforming is used to obtain the original part of that signal from converted coefficients, is called the Inverse Discrete Cosine Transform (IDCT).


The converted coefficient contains the same information as that in the original signal. The order of values obtained by applying the DCT is coincidentally from lowest to highest frequency.

the DCT is coincidentally from lowest to highest frequency. Fig. 7. Results of compression algorithm (from

Fig. 7. Results of compression algorithm (from left to right) original image, compression factor 2, compression factor 4, compression factor 8.



This paper discusses ROI-based medical image compression. IWT is recommended for medical image applications because of the perfect reconstruction with low computation complexity. Different techniques can be used for non-ROI compression of medical images. Non-ROI part must be encoded, because it gives the accurate position of ROI. ROI based coding is used along with compression for non- ROI reflects an accurate measure of performance. ROI based compression provides better performance compared with other methods. The proposed technique is less complexity and allows progressive transmission in telemedicine applications.


The authors would like to thank the reviewers of the paper for their valuable comments that helped to improve this paper.


[1]. Miaou, S.-G., Ke, F.-S., Chen, S.-C., “A lossless compression method for medical image sequences using

JPEG-LS and interframe coding,IEEE Trans. Inf. Technol. Biomed., 2009, 13, (5), pp. 818821. [2]. Placidi, G., Adaptive compression algorithm from projections: application on medical grayscale images,J. Comput. Biol. Med., 2009, 39, pp. 993999. [3]. Baeza, I., Verdoy, A., “ROI-based procedures for progressive transmission of digital images: a comparison,J. Math. Comput. Model, 2009, 50, pp.


[4]. Thomas Pintaric., An Adaptive Thresholding Algorithm for the Augmented Reality Toolkit.” Vienna University of Technology.

[5]. Sanchez, V., Abugharbieh, R., Nasiopoulos, P. Symmetry-based scalable lossless compression of 3D medical image data,IEEE Trans. Med. Imaging, 2009, 28, (7), pp. 10621071.

[6]. Maglogiannis, I., Kormentzas, G. Wavelet-based compression with ROI coding support for mobile access to DICOM images over heterogeneous radio networks,Trans. Inf. Technol. Biomed., 2009, 13, (4), pp. 458466. [7]. Doukas, C., Maglogiannis, I. Region of interest coding techniques for medical image compression,IEEE Eng. Med. Biol. Mag., 2007, 26, (5), pp. 2935. [8]. Zhang, Q., Xiaio, X. Extracting regions of interest in biomedical images,Int. Seminar on Future BioMedical Information Engineering, 2008, pp.36. [9]. Mancas, M., Gossellin, B., Macq, B. Segmentation using a region growing thresholding,” Proc. SPIE 5672SPIE, Image Processing: Algorithms and Systems, USA, 2006, pp. 388398 [10].Calderbank, A.R., Daubechies, I., Sweldens, W., Yeo, B.L. Wavelet transforms that map integers to integers,Appl. Comput. Harmon.Anal.,.1998, 5, pp. 332369. [11].Reichel, J., Menegaz, G., Nadenau, M.J., Kunt, M. Integer wavelet transform for embedded lossy to lossless image compression,” IEEE Trans. Image Process., 2001,10, (3), pp. 383392. [12]. Lehtinen, J. Limiting distortion of a wavelet image codec,” J. Acta Cybern., 1999,14, pp. 3413564. [13].Langdon,G., Gulati,A., Seiler,E., “On the JPEG model for Lossless image compression,” In 2nd Data Compression Conference, pages 172180, 1992. [14].Weinberger,M.J., Seroussi,G. and Sapiro,G., The LOCO-I Lossless image compression algorithm:

Principles and standardization into JPEGLS,” IEEE Trans. Image Process., vol.9, no. 8, pp. 13091324, Aug.


[15].ISO. JPEG2000 image coding system,” ISO/IEC FCD 15444-1, JPEG2000 Part I Final Committee Draft Version 1.0, 2000.