| Summary: | 碩士 === 國立臺中技術學院 === 多媒體設計研究所 === 94 === As the internet becomes more and more popular, there is also an increased use of digital
images on it. However, protecting intellectual property rights and keeping tracks of
end-users with rights to use similar copies of a digital image have become extremely
important and urgent. To prevent piracies of images, finger printing techniques have been
applied to images before they are being distributed. The fingerprint can be a unique ID that
will be hidden in each image. The fingerprint can be retrieved later on to prove the
ownership of the image or who the legal users are. However users could collude to tamper
with the images to generate copies of tampered images for illegal purposes. One collusion
method is to take parts from different legal copies and pieced them together to form a
newly colluded image. That is the colluded image contained parts from several authentic
fingerprinted images. Another kind of collusion is to average each overlapping pixels from
several legal copies to create a colluded copy. The hidden fingerprints would be corrupted
in the colluded images from being made up of parts from several other fingerprint images
or by averaging, respectively. In this thesis, the proposed fingerprint technique applies the
discrete cosine wavelet transform on the original image whereby a gray fingerprint image
would be hidden in the transformed domain. The gray fingerprint image consists of a
unique ID code. During recovery, Hamming code is used to improve the quality of the
recovered fingerprint ID in the event it is corrupted by collusion. Experimental results
show that the fingerprint technique is resistant to collusion attacks. In fact the IDs of the
colluders could be identified and used to prove the rightful ownership of the image and also
to pinpoint the colluders.
|
|---|
