Research Of The Lossless Compression In Medical Images Based On The Improved SPIHT Method

With the rapid development of multimedia technology in recent years, the digital image transmission plays a very important role in communication fields. Digital images generally include a great deal of information, so the compression technology becomes the key of the research. On this occasion, it inevitably becomes one of the main tasks to seek after the effective image coding algorithm. Although lossy compression method can get a higher compression ratio than lossless compression, it lacks the ability to completely restore the original image, resulting in useful details of medical image information losing and brought diagnostic difficulties or legal disputes. Therefore, medical image compression algorithm is mainly focused on the lossless compression.The theory of wavelet transform is a new math branch developed in recent years. Because of its special characteristics in space-frequency, it has become one of the main research directions in image compression coding. This algorithm is a new image coding algorithm after the Fourier transform and the DCT transform. It divides the image into several subbands in different resolution and different direction, which are compatible with human visual characters. Besides, it can split image energy into some subbands so as to code, and make transform to the whole image, which is better than the DCT transform. The traditional image compression based on wavelet transform coding, wavelet transform coefficients are floating-point type, and the process of quantify inevitably produces different degrees of distortion because of computer rounding errors of floating point operations, so it cannot be directly used for lossless image restructure. Sweldens proposed a method of wavelet construction——lifting scheme in 1994, which inherits the traditional multi-resolution wavelet transform characteristics, not dependent on the Fourier transform, and wavelet transform coefficients are integers. Sweldens has been proven to enhance the basis of an integer can be set to the integer set of wavelet transform, which can achieve lossless compression.Many scholars do a lot of work in the use of wavelet transforms for image compression and achieved considerable results, which Shaprio's embedded zero tree wavelet(EZW) coding and SPIHT algorithm are the most influential methods of wavelet image coding. In 1993, J.M.Shapiro presented the Embedded Zerotree Wavelet coding algorithm by using self-similarity between the wavelet coefficients. EZW has good MSE performance, moderate complexity and embedded bit stream, causing a compression of community interest. In 1996, A.Said and W.A.Pearlman proposed Set Partitioning in Hierarchical Tree(SPIHT), which is the improved EZW algorithm, but with higher compression efficiency, and at the same bit rate has always made superior performance to EZW, even if no need to link entropy coding. SPIHT is the second generation of embedded image coding. The basic principle of embedded coding is that the encoder sort the bite stream, which is to be coded, according to its different importance and bite stream can be truncated at any time depending on the target rate and distortion requirements, by monitoring the code of some parameters. Similarly, the decoder can decode at any time, and may be appropriate for image reconstruction. Compared to conventional DCT, embedded coding can effectively overcome the blocking effect, and get a better recovery image at low bit rates. So it has a wide range of applications in the network and the wireless transmission.Four Dimensional Medical images are arranged in chronological order by three dimensional medical images, that is, additional one-dimensional time variable on the basis of coordinate in three dimensional imaging. It can be created by all kinds of large-scale digital medical imaging devices, such as 64-slice CT, fMRI, PET,3-D ultrasound and digital subtraction C-arm machine.4-D medical images provide a clinical three dimensional, dynamic, full-depth multi-dimensional diagnostic mode. It play an important role in the areas of the treatment process planning, supporting cardiovascular surgical intervention(DSA) and quantitative analysis of functional brain activity(fMRI) et.. Although computer storage hardware cost rising and network communication bandwidth increased, the 4-D medical image data continued to show exponential growth, therefore, how full and efficient use of limited storage space and network communication bandwidth is still dependent on medical image compression method. The improved 4D-SPIHT Algorithm based on the asymmetric wavelet tree is proposed in this paper. In the scheme, a 4D asymmetric wavelet tree is used, in which way different numbers of wavelet decompositions in each dimension can be chosen. Then the paper uses SPIHT to code wavelet coefficients band by band, which speeds up the encoding speed. Experiments show that the algorithm can remove the relevant in the 3D volume data, and increased the compression performance and coding efficiency with no significant complexity of the algorithm increased.In practice, sometimes people tend to be interested in certain parts of the image, which is generally correspond to the image in a specific and unique nature of the region, and often referred to as regions of interest(ROI) and the rest as background. The lesions in medical images often referred to as ROI. The interest region of interest and non interest will be compressed differently according to observers on a particular region of the high and low level of interest, in order to reduce the total amount of data. This can not only save network bandwidth and storage space, but also to maximize the image quality to meet the requirements of the observer. In this context, ROI image compression algorithm regional priorities emerged. The current ROI, which based on JPEG2000 image compression, has proposed two algorithms: Max Shift and Scaling-based Shift, prior assigning the limited rate to the image region of interest, and improving image subjective quality of experience. The aim of segmentation in image compression is not object identification or analysis of features, but to group spatially connected pixels lying within a small dynamic gray level range. The image was divided into several regions, and then to be dealt according to different need. To achieve this objective, this paper proposes a SPIHT lossless compression based on region of interest. The basic idea is:First of all, we segmented a image with simple but efficient region growing procedure. After that, a discontinuity index image data part, an error image data part and a high-bits seed data part will be obtained. The first part, which contains scan pattern, is then algorithm coded, and the second data part is SPIHT coded. The results showed that to some extent this method affected the image compression effect after introducing the segmentation. But the extraction region of interest, compression block and block recovery can be achieved.