Image analysis – Image compression or coding – Pyramid – hierarchy – or tree structure
Reexamination Certificate
1999-09-27
2002-07-16
Johns, Andrew W. (Department: 2621)
Image analysis
Image compression or coding
Pyramid, hierarchy, or tree structure
C382S261000
Reexamination Certificate
active
06421467
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to data compression, transmission and high fidelity reconstruction of such data, especially image data. Although other information types may benefit from the present invention, it applies with special interest to complex images such as medical and satellite images.
BACKGROUND OF THE INVENTION
Large volumes of medical imagery demand high fidelity image compression for archiving and transmission. Lossy compression of medical images, however, has always been a subject of controversy since the loss of information involved is thought to affect diagnostic judgment. Most radiologists prefer not to exceed a compression ratio of 10:1 when JPEG compression is used.
Application of Wavelet Compression to digitized Radiographs
, by M. A. Goldberg, et al., Vol. 163, pp463-468, AJR(1994) describes wavelet based compression techniques which may not reduce the image quality even when subjected to compression ratio up to 30:1. Wavelet coding is a class of sub-band coding that allows selective coding of sub-bands significant to visual perception. The encoding process actually determines the distortion introduced by quantization. It is well known from rate distortion theory that vector quantization yields optimal distortion. Implementation of vector quantization, specifically the encoding process, is, however, quite involved. However, the decoding process in vector quantization is relatively simple and can be achieved from look up tables. Vector quantization of standard images is a well-researched topic. See special Issue on Vector Quantization, IEEE Trans. Image Process (1996). Other prior art approaches, in general, use a simple statistical clustering technique for generating the code vectors in vector quantization. Limitations of these earlier algorithms include long search processes and getting trapped in local minima. Yet other research discloses clustering techniques for efficient codebook design, including fuzzy clustering in deterministic annealing associated with the materials sciences.
Multi-resolution codebook design, as known in the art, provides a database of the centroid vectors or code vectors of differing resolutions. These centroid vectors as discussed herein are used to reconstruct the entire set of vectors in a given cluster within a given optimization criteria. The codebook is indexed in some ordinary fashion as known in the art. When a specific code vector is to be sent, the index into the codebook is sent instead. The receiver indexes into the same codebook and retrieves the code vector being sent. In this way the data being sent is compressed.
However, it would be desirable to provide a system that provides for image data compression and storage and/or transmission of that compressed data and subsequent high fidelity reconstruction of the image.
It is an object of the present invention to compress data by combining vector quantization and adaptive fuzzy leader clustering of the data, subsequent transmission or storage of that compressed data, followed by reconstruction of the original data.
It is another object of the present invention to provide a process for fast high fidelity transmission of images.
SUMMARY OF THE INVENTION
The above objects are satisfied and other limitations of the prior art are overcome in the present invention. The present invention includes in combination an Adaptive Fuzzy Leader Clustering or AFLC for on-line generation of code vectors, and applying AFLC to quantization of wavelet transformed sub images. The present invention may also be referred to with the acronym AFLC-VQ. The selection of the number of clusters generated is decided by the quantization level to be applied i.e. whether coarse or fine quantization is needed. AFLC-VQ is based on an integration of self-organizing neural networks with fuzzy membership values of data samples for upgrading the centroid values of the clusters as the samples are presented thus eliminating most misclassifications. This technique is also noise tolerant and avoids many problems associated with earlier vector quantization efforts particularly for large medical images. The improved performance of this new technique is its capability to generate high fidelity reconstructed images even at very low bit rate in the compressed image with acceptable MSE (mean square error) and PSNR (peak signal to noise ratio).
Vector quantization algorithms provide advantages for high fidelity transmission of medical images at very low bit rates in a fast, progressive manner for applications in telemedicine and other interactive web based downloading of large medical images in monochrome and color. In a preferred embodiment, the addition of an adaptive arithmetic coder further improves the performance of the application of the AFLC-VQ algorithm. An advantage of being able to use such low bit rate, i.e. a high compression ratio without introducing significant distortion in the reconstructed image, is that under medical emergency conditions the compressed image can be sent to a remote location quickly (since there are few bits) over the internet for immediate evaluation while images with better quality can follow in a progressive manner if desired.
A paper relevant to the present invention, authored by the present inventor, Sunanda D. Mitra, along with Shuyu Yang (who is not an inventor of the present invention), entitled High Fidelity Adaptive Vector Quantization at very low bit rates for Progressive Transmission of Radiographic Images, was published in Journal of Electronic Imaging, Jan. 1999, Vol. 8(1)/1. This paper is hereby incorporated herein by reference.
For a better understanding of the present invention, together with other and further objects thereof, reference is made to the accompanying drawings and detailed description and its scope will be pointed out in the appended claims.
REFERENCES:
patent: 5745758 (1998-04-01), Shaw et al.
patent: 5892549 (1999-04-01), Feng
patent: 5892847 (1999-04-01), Johnson
patent: 5903673 (1999-05-01), Wang et al.
patent: 6091777 (2000-07-01), Guetz et al.
Azarian Seyed
Cohen Jerry
Erlich Jacob N.
Johns Andrew W.
Perkins Smith & Cohen LLP
LandOfFree
Adaptive vector quantization/quantizer does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptive vector quantization/quantizer, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptive vector quantization/quantizer will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2899983