Computers in Biology and Medicine 39 (2009) 993--999 Contents lists available at ScienceDirect Computers in Biology and Medicine journal homepage: www.elsevier.com/locate/cbm Adaptive compression algorithm from projections: Application on medical greyscale images Giuseppe Placidi ∗ INFM, c/o Department of Health Sciences, University of L'Aquila, Via Vetoio 10, 67010 Coppito L'Aquila, Italy ARTICLE INFO ABSTRACT Article history: Received 25 June 2008 Accepted 28 July 2009 Keywords: Image compression Projection entropy Reconstruction from projections Medical imaging Medical tomography Adaptive algorithm Greyscale images Image compression plays a crucial role in medical imaging, allowing efficient manipulation, storage, and transmission of binary, grey-scale, or colour images. Nevertheless, in medical applications the need to conserve the diagnostic validity of the image requires the use of lossless compression methods, producing low compression factors. In this paper, a novel near-lossless compression algorithm from projections, which almost eliminates both redundant information and noise from a greyscale image while retaining all relevant structures and producing high compression factors, is proposed. The algorithm is tested on experimental medical greyscale images from different modalities and different body districts and results are reported. © 2009 Elsevier Ltd. All rights reserved. 1. Introduction In recent years, the need for image compression has grown steadily and it is a very important aspect of image storage and trans- mission. For example, image compression has been and continues to be crucial to the growth of multimedia computing (that is, the use of digital computers in printing and publishing and also in video production and dissemination) [1,2]. In addition, it is the natural technology for handling the increased spatial resolutions of today's imaging sensors and evolving broadcast television standards [3,4]. Furthermore, image compression plays a crucial role in medical imaging, allowing efficient manipulation, storage, and transmission of binary, grey-scale, or colour images [5,6]. A large amount of data is often produced when a series of two- dimensional (2-D) functions are sampled and quantized to create digital images or 3-D reconstructions. In fact, the amount of data generated may be so large that it results in impractical storage, pro- cessing, and communication requirements. Image compression addresses the problem by reducing the amount of data required to represent a digital image. The underly- ing basis in the reduction process is the removal of redundant data by transforming a 2-D pixel array into statistically uncorrelated data. At some later time, the compressed image is decompressed to ∗ Tel.: +39 862 433493; fax: +39 862 433433. E-mail addresses: giuseppe.placidi@univaq.it, giuseppe.placidi@cc.univaq.it (G. Placidi) URL: http://www.giuseppeplacidi.org. 0010-4825/$ - see front matter © 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiomed.2009.07.013 reconstruct the original image or an approximation of it. The com- pression techniques can be separated into two classes: lossless methods and lossy methods. The first class is composed of those methods which reconstruct an image identical to the original, the second comprises compression methods which lose some image details after their application: the reconstruction is an approxi- mation of the original image. Lossless methods seldom achieve a compression factor > 5, while lossy methods can also produce a compression factor of 100 of the same image. These differences are mainly due to the fact that lossless methods are based on the elimination of coding redundancy, while lossy methods are based on some other heuristics to obtain an approximate result. In the medical field in particular, the need to conserve the di- agnostic validity of the image requires the use of lossless compres- sion methods. A complete survey of up to date compression/coding schemes, also applied in medical imaging, can be found in [7]. A medical image contains both useful and useless information, the latter being noise, as a result of the experimental acquisition and/or reconstruction process, and it can be eliminated. Moreover, it contains symmetries and repeated patterns which can also be eliminated without degrading the image. A compression method which is able to compress and reconstruct a medical image by rejecting noise information or eliminating re- dundancies has to be considered near-lossless because it maintains useful information. We present a near-lossless compression algorithm based on a mathematic tool extensively used in medical image tomography: reconstruction from projections, RP [8–10]. RP was the first ac- quisition/reconstruction technique employed and is based on the