Cryptography Reference
In-Depth Information
image samples with some additional patterns such as bar code, words pattern,
and logo mark, etc. The realization of Image Recognition Methods will be
more challenging work compared with the Words Recognition Methods. The
mark recognition methods should be used and more image data should be
processed. This is because in order to recognize an image, more complicated
techniques are necessary in this case.
Image Feature Extraction Method:
A proposed image recognition method is called as MiLuLa [17], in which the
URL is retrieved by the method of image features extraction. As principle, let
us consider a pair of pictures showed in Fig. 10.12.
Fig. 10.12.
Image (a) and (b) are an original image and an image with severe
geometrical deformation which was taken by a mobile phone camera.
In Fig. 10.12, the images (a) and (b) are an original picture and a picture
with severe geometrical deformation that was taken using a mobile phone
camera. Generally, it is di
cult for a watermarking method to extract an
embedded watermark from an image with such severe deformation as in
Fig. 10.12(b). With the human eye, we can easily recognize that these two
images are the same pictures. Why is this? The reason is that, although the
image is severely deformed, the fundamental features in the image remain
unchanged. With these fundamental features we can recognize the image
correctly at a short time.
