Image analysis – Applications
Reexamination Certificate
1999-01-29
2001-10-09
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
Reexamination Certificate
active
06301368
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the field of image processing. More specifically, this invention relates to hiding information in compressed images, particularly for the use in wavelet-based compressed fingerprint images.
BACKGROUND OF THE INVENTION
Image understanding systems require an input image to process. The images are sensed either locally or remotely as the first step. In an unattended operation of the system (local or remote), it is possible to intercept the input image and reuse it for unknown purposes in the future. Alternatively, images can be constructed through other means or can be acquired from other sensors. In order to ensure that these kind of events do not take place, the system needs to authenticate the input image. If the input image can hide a message only known to the other stages of the system, then attacks outlined above will have no imapct as other stages can reject an input image not meeting the specifications. With the growth of the Internet over the last few years, many commercial applications are being explored. Such systems are often remotely operated and unattended. In such systems, if multimedia signals are involved for any purpose, the system must validate the signal before using it for any purpose. For example, an e-commerce system using a fingerprint of the subject to validate a transaction over the web such as airline ticket purchase needs to ensure that the fingerprint image being transmitted from the remote client is not an earlier acquired image or otherwise constructed image. By hiding a new message string in the image every time the image is acquired, the system can make sure that the image is not a stale old copy. If the message is hidden in known places in the image, it is easy to modify the image to the desired message. Often because of their large size, the images are compressed using either domain specific or general compression techniques to save bandwidth. The problems illustrated earlier apply to these compressed images as long as the compression standard is known.
Fingerprints have been used for authentication and identification purposes for several decades. A typical automatic fingerprint identification systems consists of an image acquisition stage (
110
) followed by two other stages as shown in FIG.
1
. Fingerprint matching comprises two steps: feature extraction (
120
) and feature matching (
130
). The following reference describes examples of the state of the prior art:
N. K. Ratha, S. Chen and A. K. Jain, Adaptive flow orientation based feature extraction in fingerprint images, Pattern Recognition, vol. 28, no. 11, pp. 1657-1672, November 1995.
This reference is incorporated herein by reference in its entirety.
In step
110
, the image is acquired. This acquisition can be either local or remote. There are several techniques available for sensing the image. These include optical, capacitance, thermal and ultrasound. Depending on the application, either the sensed image can be locally acquired or remotely acquired. For example, in a home-based banking system allowing remote access to a user's bank account based on fingerprint-based authentication will require the fingerprint to be sensed on the user's fingerprint scanner located in his home. Other examples of remote fingerprint acquisition include point of sale transaction authorization based on fingerprints. In such situations, verification can be done locally or the image can be compressed and transmitted to a server over a network.
Image compression techniques in general try to eliminate redundant information in the image description thereby achieving a compact representation of the signal. Lossless compression schemes usually offer low compression ratios while retaining the ability to fully reconstruct the image whereas lossy compression techniques offer very high compression ratios with degradation of the visual image quality. There are a number of standard as well as domain specific techniques available to compress images. Standard compression techniques include JPEG and GIF. Many other domain specific compression algorithms have been proposed in the literature to compress grayscale fingerprint images because the standard techniques have a tendency to blur the high frequency structural ridge features in fingerprint images. Thus, the decompressed image has lost the necessary detail which is needed for further processing. Wavelets offer a compact representation scheme for signal representations. Many wavelet-based signal and image processing techniques have been described in the literature. Recently, the FBI has proposed a fingerprint image compression and decompression standard known as the Wavelet Scalar Quantization (WSQ). The following reference describes the standard.
WSQ Gary-scale fingerprint image compression specification IAFIS-IC-011v2 (rev 2.0), February 1993.
Drafted by T. Hopper, C. Brislawn, and J. Bradley, Federal Bureau of Investigation
This reference is incorporated herein by reference in its entirety.
Data hiding is a form of steganography to embed data (messages) into digital media for the purpose of identification, annotation and copyright. Data hiding techniques ensure that embedded data remain inviolate and recoverable while not restricting the access to the media. The size of the message that can be efficiently hidden is limited by many factors including the image size. One of the main application of data hiding techniques is to embed annotations about the media such as features of an image including description and name or any other information the user feels important for future. Yet another application of data hiding is the placement of a digital watermark. Various techniques for data hiding are surveyed in the following reference.
W. Bender, D. Gruhl, N. Morimoto and A. Lu, Techniques for data hiding, IBM Systems Journal, Vol. 35, No. 3 &4, 1996.
This reference is incorporated herein by reference in its entirety.
PROBLEMS WITH THE PRIOR ART
Data hiding in a compressed signal poses several challenges. First, the possible places to hide the message have been shrunk significantly by the compression process. Secondly, any alteration of the compressed coefficients or indices will result in significant changes in a group of pixels depending on the compression scheme used. For biometrics images, there are extra challenges. If the modifications made in the compressed coefficients cause changes in the structural properties of the image or any other feature being used by subsequent processes, the distortions after decompression are not tolerable. If possible, the perfect reconstruction should be supported. Yet another problem with prior art is that one can use predictable sites to hide the message. This has the problem that once it is known by hackers, the hidden message can be easily recovered.
Sandford et al. (U.S. Pat. No. 5,727,092) describe a system wherein a message is embedded in the coefficients of a WSQ compressed image. However, first Sanford requires that the decoder be supplied a separate key in order to extract the hidden message (page 5 lines 22-25, page 8 lines 11-41). Transmitting this key in a secure way then becomes a concern. Second, in Sandford, sites are chosen more or less without particular regard to the effect their modification has on the compressed image. Their process starts by aggregating the coefficients from all scales into a single histogram (page 4 lines 33-40) which is then used for site selection. The only limitation is that the values must differ by only a small amount (typically one). Third, Sandford scans the coefficient file sequentially from the beginning until he finds a value that matches one of his candidate pairs. He then modifies this site and continues scanning from where he left off (page 7 lines 29-37). This means that, at least for short messages, the altered sites are all near the beginning of the coefficient file and the embedded bits are in the same order as in the message itself.
The following references give further background and are incorporated by reference in thei
Bolle Rudolf Maarten
Connell Jonathan Hudson
Ratha Nalini Kanta
International Business Machines - Corporation
Johns Andrew W.
Nakhjavan Shervin
Percello Louis J.
LandOfFree
System and method for data hiding in compressed fingerprint... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for data hiding in compressed fingerprint..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for data hiding in compressed fingerprint... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2560416