The Study Of Image Compression Method For PACS

With the improvement of medical imaging, image technology, computer technology and net communication technology, a novel picture archiving & communication system-PACS that can be used to the sampling, storage, management and transmission of medical image, has developed recently. The goal function of PACS is to transfer all medical images resulting from all sorts of medical imaging instruments (such as X-ray imaging, molecule imaging, CT, MRI, and B-mode ultrasound, et. al) to digital images, which will be transmitted to center-image management system. In the same time, the medical images will be transmitted to image displaying workstation for clinical doctors. PACS not only can realize the electronic medical images(non-film), but also improved the efficiency of the transmission, retrieval, searching and sharing for medical images. Because of the huge data, medical image must be compressed before storage and transmission; on other hand, the image compression technology has been made much progress recently. However, the image compression method for PACS must fit the properties of PACS application. Therefore, it is very necessary to study the encoding algorithm for medical image transmission, storage and retrieval in PACS.Firstly, based on studying the theory of wavelet transform and properties image signal, the author described a para-symmetry boundary extension method and realized 2-D discrete wavelet transform by means of 1-D wavelet transform according to correlation of adjacent pixels. Also, the author proved that the discrete wavelet transform of inverse data stream in row and the sign of discrete detail signal will be inversed, and constructed a filter of wavelet transform based on symmetry of bi-orthogonal filter. Then, the 2-D image transform can be realized easily.Secondly, the author has researched the wavelet coefficients distribution property based on the histgram of detail subimages, and proposed an adaptive image quantization algorithm according to local data distribution in every subimage. In this quantization procession, every subimage will be disparted to some sub-blocks, and the sub-blocks will be quantized with different quantizer according the coefficients distribution; when the bias is large, the sub-block will be quantized with high accuracy; when the bias is little, the sub-block will be quantized coarsely. Consequently, according to the weight of different frequency detail subimage in wavelet image reconstruction, the author described a mixed 2-D gray image encoding method, which encode the low frequency analyze subimage with JPEG standard because of its significance, and encode thequantization results with run-length coding and Huffman coding. The test has proved that, in the same sub-block mode, center bias threshold is higher, the compression ration is higher and SNR of reconstructed is lower; on other hand, for the same center bias threshold, the dimension of sub-block is larger, the compression ration is higher and SNR of reconstructed is lower too.Thirdly, according to the need of PACS, the author proposed a mixed-encoding-algorithm-based multi-resolution encoding scheme and region-of-interest encoding scheme. The former makes use of the multi-resolution analyze property of wavelet transform, and can accelerate medical image retrieval. The latter makes use of space corresponding property in every wavelet subimages, and can increase the total medical image compression efficiency.As a whole, in this thesis, the author has studied the 2-D image wavelet transform and proposed a mixed image encoding scheme that is fit to PACS. Also, the author has designed multi-resolution encoding scheme and region-of-interest encoding scheme according to 2-D wavelet transform property. However, there are much work to do for this research practical application in PACS, such as the selection of optimized center bias threshold and combination of vector quantization.