Image analysis – Image compression or coding – Lossless compression
Reexamination Certificate
1998-12-21
2001-07-31
Tran, Phuoc (Department: 2721)
Image analysis
Image compression or coding
Lossless compression
C382S251000
Reexamination Certificate
active
06269193
ABSTRACT:
FIELD OF THE INVENTION
This invention relates in general to digital medical imaging, and in particular to the compression of projection radiographic images that have been digitally acquired.
BACKGROUND OF THE INVENTION
Image compression is essential to the successful deployment of digital medical imaging systems, also referred to as Picture Archive and Communications Systems (PACS). Image compression is a processing operation that reduces the amount of digital data required to represent an image, and hence, enables more efficient utilization of both available network bandwidth and image archive storage space, thereby reducing the cost to implement PACS. There are fundamentally two types of image compression. The first type is known as lossless compression. With lossless compression the reconstructed image identically matches the original and is therefore a fully reversible process. Since the diagnostic quality of the compressed and reconstructed image is assured relative to the original image, lossless compression is appealing for medical applications. The major drawback of lossless compression is the limited compression ratios that can be achieved, which is typically on the order of 2:1, or equivalently a 50% reduction in file size. Lossy methods can achieve much greater compression ratios, on the order of 10:1 or higher, and can therefore provide more significant cost savings for PACS implementations. However the medical community has been slow to accept lossy compression for fear that important diagnostic information could be lost. It is therefore necessary to provide a means of achieving higher compression ratios than those achievable with lossless methods but in a manner that will be widely acceptable to the medical imaging community. Such a method would provide reasonable cost savings for PACS and minimize the potential for loss of important diagnostic information.
Reid et al, “Second-generation image coding: an overview,” ACM Computing Surveys, Vol. 29, No. 1, pp. 2-29, March 1997, in overviewed recently-developed image compression techniques that have been termed second-generation image coding. These methods incorporate properties of the human visual system into the coding strategy in order to achieve high compression ratios while maintaining acceptable image quality. The techniques utilized in second-generation image coding are based on visual patterns, multi-scale decomposition, contour-coding, and segmentation. Visual-pattern based approaches use the fact that the eye can decompose the overall image into a set of smaller “visual patterns”. Multi-scale decomposition techniques create sets of progressively smaller images and identify common features in the image that are present at the various levels of detail. All second-generation coding techniques are lossy in nature. However, these methods attempt to identify and separate visually significant and visually insignificant areas of the image, and apply appropriate coding techniques to each area.
Researchers in W. C. Chang, et al, “Lossless image compression methods for PET imaging,” Biomedical Engineering-Applications, Basis & Communications, Vol. 8, No. 3, pp. 309-316, June 1996; L. Shen et al., “Segmentation-based lossless coding of medical images,” Proceedings of the SPIE-The International Society for Optical Engineering Conference, 24-26 May 1995, Taipei, Taiwan, SPIE, Vol. 2501, pp. 974-982; and V. Vlahakis et al., “ROI approach to wavelet-based, hybrid compression of MR images,” Proceedings of 6
th
International Conference on Image Processing and its Applications, 14-17 Jul. 1997, Dublin, Ireland, Part Vol. 2, pp. 833-837 proposed segmentation-based hybrid lossless image compression coding methods for medical images. W. C. Chang et al, “Lossless image compression methods for PET imaging,” Biomedical Engineering-Applications, Basis & Communications, Vol. 8, No. 3, pp. 309-316, June 1996, described a hybrid lossless coding method for PET (Positron Emission Tomography) images. The supported region (cross-section region) and unsupported region (background region) was separated by a binary mask using a thresholding segmentation algorithm. The unsupported region was not encoded while the supported region was encoded using a lossless entropy coding method. However, the boundary of the binary mask, which is the contour of the segmented supported region, had to be encoded using the chain code method. Extra bytes for describing the shape of the contour also needed to be provided to both encoder and decoder in order to reconstruct the image.
Another segmentation-based lossless coding method was applied to digitized mammography and chest radiography film images in L. Shen et al., “Segmentation-based lossless coding of medical images,” Proceedings of the SPIE—The International Society for Optical Engineering Conference, 24-26 May 1995, Taipei, Taiwan, SPIE Vol. 2501, pp. 974-982. The region growing scheme was used to generate segments at which gradients of gray levels were within certain thresholds. The discontinuity index map data sets were also generated to present the pixels which separated segments. An entropy coding method was applied to code segments individually. However, it was necessary to send extra data with the compressed image to correctly index the segments for decompression.
One region of interest (ROI) approach to a wavelet-based hybrid compression method for magnetic resonance (MR) images was proposed in V. Vlahakis, et al., “ROI approach to wavelet-based, hybrid compression of MR images,” Proceedings of 6
th
International Conference on Image Processing and its Applications, 14-17 July 1997, Dublin, Ireland, Part Vol. 2, pp. 833-837. A MR image was decomposed using a wavelet transform into three scales. Starting at the middle scale, scale
2
, the radiologist clicked the mouse at a seed pixel inside the area which he identified as the ROI (corresponding to brain tissue, tumors, etc.), and a seed-fill segmentation algorithm scanned and labeled the pixels around it until a boundary was detected. Then fine or no quantization was used for wavelet coefficients corresponding to ROI, and coarse quantization for the rest of the coefficients of scale
1
and
2
. Finally, the quantized coefficients were run-length coded and the resulting run-lengths were compressed with a Huffman code. The low-pass residue of scale
3
was losslessly compressed using DPCM (Differential Pulse Code Modulation).
Other hybrid lossy and lossless compression methods for consumer images are disclosed in the following U.S. patents. U.S. Pat. No. 5,552,898, issued Sep. 3, 1996, to inventor F. A. Deschuytere teaches a method of lossy and lossless compression in a raster image processor on output devices. Digital input commands defined in a page description language are separated into two types of instructions. The first set of instructions comprise solid regions on the printed output, which are filled with recorder elements (e.g., ink) of the same highest or lowest density value, and second instructions resulting in halftoned regions, which are getting different densities. It is advantageous to distinguish a first type of instructions from a second type of instructions, and treat them separately. The information stored in the first type of instructions is compressed by a lossless compression method (recommended by CCITT—International Telegraph and Telephone Consultative Committee). As such, solid patterns will appear on the rasterized image at the highest resolution and without any quality loss. The second type of instructions on the other hand corresponds to continuous tone image or intermediate tone graphical information. A slight deterioration of the information contents is acceptable and will be hardly noticeable, thus is compressed by a lossy compression method (JPEG). When all the digital input commands for one page are handled, the compressed data can be retrieved and combined to reconstruct the rasterized image.
U.S. Pat. No. 5,553,160, issued Sep. 3, 1996, to inventor B. J. Dawson teaches a method and apparatus for dynamically sele
Foos David H.
Whiting Bruce R.
Young Susan S.
Alavi Amir
Eastman Kodak Company
Noval William F.
Tran Phuoc
LandOfFree
Method for statistically lossless compression of digital... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for statistically lossless compression of digital..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for statistically lossless compression of digital... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2472907