Image processing algorithms for security applications. Yesna Oyku Yildiz

ISBN: 9780549523628

Published:

NOOKstudy eTextbook

102 pages


Description

Image processing algorithms for security applications.  by  Yesna Oyku Yildiz

Image processing algorithms for security applications. by Yesna Oyku Yildiz
| NOOKstudy eTextbook | PDF, EPUB, FB2, DjVu, audiobook, mp3, ZIP | 102 pages | ISBN: 9780549523628 | 6.15 Mb

Image processing may be defined as the intentional manipulation of the image function in order to achieve various goals. Image manipulation from a perception point of view could be classified into two groups---mainly those image manipulations whichMoreImage processing may be defined as the intentional manipulation of the image function in order to achieve various goals.

Image manipulation from a perception point of view could be classified into two groups---mainly those image manipulations which are intended to be perceived by the audience, and those manipulations which are to be invisible to the intended audience.-This thesis is founded on the study of various purposeful goals which belong to, or fall within, both of the two image manipulation classifications. The motivation for some of these image manipulation algorithms was brought about mainly by the need to provide a better and more improved screening quality for todays Automated Explosive Detection Systems.-This thesis solves two purposeful problems for each of the two classifications, mainly steganography and 3-D Threat Image Projection, both belonging to the imperceptible manipulation classification and then edge detection and image segmentation which belong to the first classification i.e.

perceivable manipulations.-The steganography problem this thesis is trying to solve is to create a high capacity and secure algorithm for JPEG images. JPEG images are chosen due to its abundant use and commonly accepted format. The algorithm shall be secure against statistical attacks in the spatial domain i.e. image data and in DCT coefficient domain. It will be secure against attacks that could use the JPEG header to determine message embedding. It will be key-dependent to further increase security against statistical attacks. One way to achieve these goals is to manipulate the JPEG compression algorithm with an arbitrary embedding scheme.

This thesis introduces a new key-dependent quantization matrix algorithm which improves data hiding security. To the authors knowledge, quantization tables have been investigated for compression purposes but not for steganography purposes. Having a key-dependent quantization table adds a level of security since only the sender and the recipient know the key and true quantization table used and eavesdroppers cannot reconstruct the true image.

Any statistical attacks using image data would be on a different image than the one with the hidden message. This makes it much harder to determine the length of the hidden message and/or extract the hidden message. Having a key-dependent table adds another level of security since each image could be created with a different table. However, to do this, the quantization table needs to compress as well as JPEG and the image needs to look like JPEG image when reconstructed. Otherwise, eavesdroppers could suspect that the image was manipulated to hide data.-Fibonacci p-codes are used to determine the key-dependent quantization table and Model-based Steganography to embed the message data.

The performance of this approach was measured using DCT histograms and image histograms on various grayscale images. The differences in histogram standard deviations between the encoded and clean images were analyzed with the various quantization tables. Results suggest that the key chosen to generate the quantization table should be specific to the image and the encoding ratio.-The second problem in imperceptible image manipulations is to come up with a proof of concept for a 3-D Threat Image Projection (TIP) algorithm that does not demonstrate contextual cues.

TIP is a software system that exposes screeners to artificial but realistic images of threat items during routine baggage screening operations[ECAC07]. There are two methods of TIP operation. Fictional Threat Image (FTI)---an image of a threat (such as a gun) that is selected from the image library and then superimposed onto a real passenger bag image and Combined Threat Image (CTI)---insertion of an entire image of a whole bag containing a threat item [ECAC07]. The CTI method is a costly method that requires a very large library of threat bags that also needs to be continuously updated in order to minimize TIP image recognition by the screeners.

CTI TIP only works for certain installations where the screeners cannot see the actual bags going through the scanner like the remote viewing systems used for checked baggage. (Abstract shortened by UMI.)



Enter the sum





Related Archive Books



Related Books


Comments

Comments for "Image processing algorithms for security applications.":


ekipacs.eu

©2008-2015 | DMCA | Contact us