Neural-network-based method of image compression

Details Neural-network-based method of image compression Neural-network-based method of image compression

2000-05-05

2004-09-28

Couso, Jose L. (Department: 2621)

Image analysis

Image compression or coding

Reexamination Certificate

active

06798914

ABSTRACT:

FIELD OF THE INVENTION
The invention relates to the field of image compression. More specifically, the invention relates to a technique for compressing images using neural-network-based artificial intelligence methods.
BACKGROUND OF THE INVENTION
Image compression techniques are useful in many applications, including picture and video communications, picture and video storage and image analysis. Many prior-art techniques exist for performing such compression.
Compression techniques fall into two general categories, lossy and lossless. In lossy compression, images may be compressed into very efficient formats. However, the cost of doing so is a loss of information incurred in performing the compression. This may result in a reconstructed image not having all the features of the original image.
In lossless compression, as the name implies, all information is retained when the image is compressed. However, in prior art techniques, the compression ratios (i.e., the ratio of the amount of data prior to compression to that after compression) achieved are generally not nearly as great as those achieved using lossy techniques. Typically, lossless compression ratios are 4:1 or less.
Prior-art image compression techniques include JPEG, wavelet-based techniques and fractal-based techniques. The problem with such techniques is that in order to get the compressed image data small enough to be of practical use, fine details are lost. While this is fine for some applications, in which fine details are not necessary, it is unacceptable for other applications, for example, remote diagnosis using x-ray images, in which all details are necessary.
It would, thus, be desirable to have a lossless compression algorithm which is also efficient (i.e., provides high compression ratios).
OBJECTS AND SUMMARY OF THE INVENTION
It is an object of the invention to provide such an algorithm, being both lossless and highly efficient.
It is an object of the invention to provide a neural-network-based technique for compressing images.
It is an object of the invention to provide a technique for compressing x-ray and other diagnostic images so that they may be reconstructed without loss of detail.
These and other objects are accomplished by the techniques of the present invention. The inventive compression method is an artificial-intelligence-based approach that uses neural networks and a look-up table (LUT). By saving the information on the edges of an image, the algorithm symbolically represents portions of the image using the LUT. The look-up table is constructed “on the fly”, as the input image data is scanned in. Symbols from the LUT, when used to represent the image, are highly repetitious, and they are compressed using an additional encoding technique, like run-length-limited (RLL) coding. The algorithm replaces repeating strings of one or more characters of an input stream with n-bit symbolic codes in an output stream. The inventive technique is capable of providing compression ratios of 100:1 or more, while allowing reliable image reconstruction.


REFERENCES:
patent: 5003299 (1991-03-01), Batson et al.
patent: 5005206 (1991-04-01), Naillon et al.
patent: 5041916 (1991-08-01), Yoshida et al.
patent: 5068644 (1991-11-01), Batson et al.
patent: 5163111 (1992-11-01), Baji et al.
patent: 5200816 (1993-04-01), Rose
patent: 5204665 (1993-04-01), Bollman et al.
patent: 5285297 (1994-02-01), Rose et al.
patent: 5313534 (1994-05-01), Burel
patent: 5343554 (1994-08-01), Koza et al.
patent: 5426745 (1995-06-01), Baji et al.
patent: 5670986 (1997-09-01), Leak
patent: 5673065 (1997-09-01), DeLeeuw
patent: 5764357 (1998-06-01), Dockser
patent: 5799111 (1998-08-01), Guissin
patent: 5812700 (1998-09-01), Fang et al.
patent: 5841904 (1998-11-01), Sugiura
patent: 5872864 (1999-02-01), Imade et al.
patent: 5898794 (1999-04-01), Sonohara et al.
patent: 6128346 (2000-10-01), Suarez et al.

Affiliated with

Also associated with

Rating

Neural-network-based method of image compression does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Neural-network-based method of image compression, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Neural-network-based method of image compression will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3232912