Page 186 - Algorithm Collections for Digital Signal Processing Applications using MATLAB

4. Selected Applications 175

9. BINARY IMAGE WATERMARKING USING
WAVELET DOMAIN OF THE AUDIO SIGNAL

The process of hiding the information like text, binary image, audio etc. into
another signal source like image, audio etc. is called watermarking. The
approach involved in watermarking the binary image signal in the wavelet
domain of the audio signal is described in the example given below.

9.1 Example

Step 1: Consider the audio signal sampled with the sampling rate of 11025 Hz.

Step 2: The audio signal is divided into frames with 1024 samples. Decompose
every frame of the speech signal using ‘db4’ wavelet transformation
[See wavelet transformation in chapter 3]. Hide the first bit of the
binary data collected from the binary image into the third level detail
co-efficients obtained using wavelet decomposition of the first audio
frame. This is done by changing the values of the third level detail co-
efficients such that mean of the third level detail co-efficients is
modified to ‘m+k’ if the binary value is ‘1’ or to ‘m-k’ if the binary
value is ‘0’.Note that variance remains same.[See Mean and variance
normalization in Chapter 2]. The variable ‘m’ is the mean of the fifth
level detail co-efficients. The value for the constant ‘k’ is chosen as
th
1/5 of the energy of the third level detail co-efficients.

Step 3: This is repeated for the other frames of the speech signal for hiding
the entire bits obtained from the binary image.

Step 4: Thus Binary image is hidden in the wavelet domain of the audio
signal.

Step 5: The hidden binary data is retrieved by comparing the mean of the
corresponding detail co-efficients computed before and after hiding.
If the mean of the third level detail co-efficient of the particular
frame corresponding to the original signal is greater than the mean
of the third level detail co-efficients of the respective frame
corresponding to the watermarked signal, the bit stored is
considered as ‘1’,otherwise ‘0’.

Step 6: Note that the duration of the audio signal is chosen such that all the
binary data collected from the binary image are hidden in the audio
signal.