Compression of CT Images with Modified Inverse Pyramidal Decomposition ROUMEN KOUNTCHEV Department of Radio Communications Technical University of Sofia Bul. Kl. Ohridsky 8, Sofia 1000 BULGARIA rkountch@tu-sofia.bg http://www.tu-sofia.bg ROUMIANA KOUNTCHEVA T&K Engineering Mladost 3 Pob 12 Sofia 1712 BULGARIA kountcheva_r@yahoo.com Abstract: - In the paper is presented one novel approach for adaptive compression of groups of computer tomography (CT) images, based on modified branched inverse pyramid decomposition. In result is obtained high compression ratio with retained image quality. To achieve this, was used the high correlation between sequences of CT images, representing same object(s). The experimental results confirm the efficiency of the presented method. Key-Words: - Image processing, Archiving of CT images, Compression of sequences of medical images; Image group coding, Modified branched pyramidal image decomposition. 1 Introduction Electronic healthcare records are important part of medical diagnostics and healthcare. As it is well- known, the old practice was based on paper rectors, which have significant disadvantages, related to legibility, resistance to humidity, accessibility, etc., etc. The contemporary approach is based on electronic rectors, which offer significant advantages, especially when visual medical data is concerned. This comprises various X-ray or ultrasound images, electrocardiograms, and many more. Computed tomography (CT) is one of the main contemporary diagnostic procedures. It uses special x-ray equipment to obtain cross-sectional pictures of the body, and for this, usually sequences of images for every patient are created and used. Important stage in CT is archiving the images obtained in an efficient manner concerning the data volume occupied and the image quality. A vast number of medical image compression techniques already exist [1-3,7], which can be divided into two large groups – lossless [4,9] and lossy [5,6] depending on the ability to restore the image fully or not. In both groups some type of image decomposition is used (for example, linear orthogonal transform or a wavelet one combined with spectral coefficients rearrangement and entropy coding). The most famous file format, used for medical images archiving is DICOM [8], which is based on the JPEG standard. In [14] is proposed one approach based on adaptive sampling of DCT coefficients. The quality of the restored images after this compression is comparable to that of the JPEG2000 as the author shows while the JPEG coder produces images with PSNR between 31 and 40 dB for same levels (i.e. visually lossless). In [5] authors confirm that wavelet decomposition assures better quality for the compressed images being than the JPEG-based coder. Some resent publications emphasize the advantages of the wavelet decomposition for medical image compression combined with other techniques in order to construct more efficient coders (for example, using joint statistical characterization [1], by linear prediction of the spectral coefficients [13], introducing region of interest (ROI) [6], incorporating planar coding [12], etc). Nevertheless the higher compression levels achieved, there is also reported significant reduction of the visual quality of these images [9,10]: while cumulative quality measures such as PSNR stay high the smoothing of vast image areas due to the wavelet coefficients quantization becomes intolerable for higher compression ratios. In this paper is proposed one new approach for lossy compression of sequences of CT images with Advances in Systems Theory, Signal Processing and Computational Science ISBN: 978-1-61804-115-9 74