Research Of Medical Image Compression In PACS

How to compress the medical image more efficiently with the specific characteristics different from general image is an urgent program in the development of PACS technique. For the advantages of supporting embedded bits flow, region of interest (ROI) coding, progressive transmission, and so on, JPEG2000 has been adopted in the extension of DICOM3.0. However, JPEG2000 does not support exact ROI coding, because it uses original discrete wavelet transform (DWT) and block based coding algorithm named EBCOT. Meanwhile, the compression of medical image sequence is not specially mentioned in JPEG2000.Aims at perform the exact ROI coding with object based image, this paper studies on shape adaptive discrete wavelet transform (SA-DWT) and applies it in image decomposition of compression. SA-DWT makes the nums of object image strictly same in original and transform domain, that leads to non-redundant coding and decoding with object image. In order to support lossy to lossless compression, integer wavelet transform with lifting schema is adopted.SPIHT is used to code the coefficients of SA-DWT. As a famous embedded wavelet zero-tree coding algorithm, SPIHT is easy to realize and modify to suit for SA-DWT based coding than EBCOT algorithm in JPEG2000. This paper proposes to modify the structure and sorting pass of original spatial orientation tree in SPIHT to improve the coding efficiency. The experiment results illustrate the modified SPEHT based on SA-DWT is not only eliminate the blur outward object border of lossless decoding contrast to that in JPEG2000, but also improved the compression rate in some level. Further more, this paper advances to move up the refinement pass one bit-plane, which improves the HVS quality of the decoding image in progressive transmission, and mostly results in higher PSNR than original SPIHT.Focus at the compression of medical image sequence which JPEG2000 does not cover, this paper studies on medical image registration and applies it as a special prediction technique (RBDP) with image sequence in order to reduce the redundant between image sequence in max. This idea offers a novel possible solution of prediction based compression of image sequence.Furthermore, this paper also discusses the design and development of a practical PACS.