Digital Watermarking of Medical Images - Proceedings of All India Seminar on Biomedical Engineering 2012 (AISOBE-2012) - page 107

Fig. 1 Frequency domain
techniques
Frequency Domain Techniques
As discussed above in these techniques, image data is transformed using some
well-known transforms like FFT, DCT, DWT etc. The coefficients of transformed
images are then coded and again inverse transformed to get a watermarked image.
Quantisation of coefficients is done prior to coding (Fig. 1 ).
More on these techniques is presented in Ref. [ 2 , 3 ].
Proposed Method
The method presented here is a digital watermarking technique which extracts
some crucial information from the image, and maps that information in the non-
ROI region of the image. The parameters required to extract the data from the non-
ROI region is embedded into the ROI of the image. The whole process is software
dependent and can be reversed by using the same piece of software used to
watermark the image. Figure 2 depicts the watermarking process.
In our method, ROI is first determined from the image and separated from the
complete image. The remaining, i.e., non-ROI region includes patient data and
other medical information. This information is then extracted from the non-ROI
region. This data are extracted on the basis of font color, which is mostly white in
DICOM images although OCR can also be used. The row and column infor-
mation is obtained from white pixels, which is scrambled and then embedded in
the non ROI region surrounding the ROI. The start and end information of these
data are then randomly mapped to the ROI region. Other than start and end,
information about where row and column is present with the information about
extant of ROI area which is also embedded in ROI region in a way that it does
not deform the image.