Developing Lossy-to-Lossless X-Ray Image Compression Using RoI Based Fuzzy C-means Vector Quantization A. D. Setiawan 1* , A. B. Suksmono 2 , B. Dabarsyah 3 , T.L.R Mengko 4 School of Electrical Engineering and Informatics - Bandung Institute of Technology 1,3 Signal and System Laboratory 2 Microwave & Radio Telecommunication Laboratory 4 Biomedical Engineering Email: 1 setiawan232@students.itb.ac.id, 2 suksmono@ltrgm.ee.itb.ac.id, 3 budiman@lss.ee.itb.ac.id, 4 tmengko@itb.ac.id X-ray image is an important part of a patient’s health history. X-ray image must be store for information retrieval or transmission in the future. The problem for storing and transmitting is the size of x-ray image itself. X-ray images come out with the size of 3 MB or even bigger. Image compression is the answer to overcome storing and transmitting problem. The compression technique must follow several requirements. The first and the most important requirement is that there must be no missing information during the compression process. This is because; the missing information could be a very important part to conduct a diagnosis process. The other requirement is that the compression technique must have the most optimal compression ratio. Lossy-to-lossless compression scheme is introduced to answer the requirements. This compression developed over vector quantization technique. Fuzzy c-means (FCM) is applied to create a set of code vector or codebook. The codebook is used for encoding and decoding process. The encoding process will result an encoded image consist of indexes including in the codebook. Meanwhile, the decoding process will result a reproduction image. The error of this reproduction image is called as image residue. Then, the encoded image and its residue will be store in the image DB or medical image for future use. Lossy-to-lossless scheme make a possibility for physicians to view lossless information over reproduction image (lossy) on a certain region of interest. * Responsible author. E-mail: setiawan232@students.itb.ac.id 1. Introduction X-ray medical image or radiology image is one of instruments for conducting diagnosis process. Storing radiology images dealing with a patient is a necessity to record patient’s medical history along with textual information. These images will be an advantage for patient as it will be a complementary data for future examination. Radiology image archiving will also give a contribution on health science research activities. Problems will arise when storing radiology image is occurred. The size of digital radiology images is big, around 3 MB or even bigger. It will need a large of storage capacity to store a huge number of radiology images. It also needs a wide bandwidth telecommunication network to transmit radiology image, especially across the net between urban and rural area. Image compression using vector quantization (VQ) is introduced to overcome those obstacles. VQ is suitable compression technique to gain fine compression ratio (12) . Fuzzy c-means is chosen to generate a good codebook compared with standard k-means clustering. VQ is a lossy compression method; meanwhile there must be no missing information in radiology image compression. The missing information during the compression process might be a very important part to do the diagnosis process. Lossy-to-lossless scheme is introduced to avoid missing information, while maintaining fine compression ratio. Lossy-to-lossless scheme is worked by enhancing reproduction image to its original image on a certain region of interest (RoI). The enhancement process is done by adding residue image and reproduction image over certain RoI. This enhancement will bring source image quality over that RoI. 2. Fuzzy Vector Quantization Image Compression 2.1 Vector Quantization (VQ) A vector quantizer Q of dimension k and size N is defined as a mapping from a vector in k-dimensional Euclidean space R k into a finite set C containing N reproduction point called as code vector or codeword. Thus, (1) C Q k → ℜ : where C = (y 1 , y 2 , …, y N | y i ∈ R k ). C is a set of code vectors and called as codebook. Fig. 1 Encoder and decoder diagram of VQ A vector quantizer is containing two component operations which are vector encoder and vector decoder. The Proceedings of the International Conference on Electrical Engineering and Informatics Institut Teknologi Bandung, Indonesia June 17-19, 2007 B-16 ISBN 978-979-16338-0-2 508