| |
Prof. Hideo Kuroda
FPT University |  |
| |
| |
|
Abstract: | | |
| |
Due to the remarkable popularization of mobile phones with built-in
high-function camera and the speedup of the network, creation of content
on personal basis and uploading and downloading of created contents are
performed frequently. In such a situation, the content of others can
also be seen freely. As a result, infringement of copyright such as
illegal copying occurs frequently irrespective of the presence or
absence of malice.
Therefore, countermeasures against copyright infringement are extremely
important. One of the countermeasures is a digital watermark. In the
digital watermarking method, the rights holder registers his own
copyright information into the certificate authority, and embeds the
copyright information in own content in an invisible form. Then,
copyright is asserted by extracting own copyright information from
currently distributed images. Generally, complicated processing is used
for embedding / extracting copyright information.
On the other hand, image compression such as JPEG or MPEG is performed
from the viewpoint of reducing the image transmission speed and
compressing the file storage memory amount. In these, the pixel value
changes by quantization and the high frequency component compression.
In addition, geometric processing such as enlargement / reduction and
rotation of images, decorative processing such as border decoration may
also be performed. As described above, the types of attacks on images
are extremely varied, and it is extremely difficult for the copyright
holder to identify and
extract their own copyright information from a large amount of images
subjected to various kinds of attacks.
In joint research by FPT University, Nagasaki University, and Kanazawa
University, we propose a candidate image retrieval method for illegal
copies without any digital watermarking technology.
By this, the copyright holder using own digital watermarking technique
may only check a quite small number of illegal copy candidate images
identified by pre-searching.
The proposed method by the universities is simple and robust for various
attacked images.
| |
|