Page 30 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

18 Chapter 1 Congruence of deep learning in biomedical engineering

5. Proposed method by the authors

In the proposed strategy we utilized discrete wavelet change
for changing a picture from its spatial domain to recurrence
space. A wavelet starts at zero and returns to zero, and therefore
is called a wave-like frequency domain. Fourier transform tech-
nique is used to construct a time-frequency representation of a
signal simultaneously. The principal reason for changing a
picture into the frequency domain during steganography is
because we embed our secret data into the frequency area
because it is hard to identify steganographically. In DWT for an
image then we separate the high frequency what is more, low-
frequency data. Low-frequency data are incorporated data about
the smoother spots of the image and are very sensitive data
where slight alteration influences the stego picture. Then again,
high-frequency information contains data on the edge, corner,
and so on of an image. Thus a change in these data results in
less noise in the reproduced picture. It is an implement that
splits up information into various frequency mechanisms; then,
evaluation of every element with determination exactly matched
to its scale is determined with the power of channels observed by
factor 2 substitute sampling. DWT is likewise invertible and can
be symmetrical. In this proposed technique we utilized Haar
wavelet change proposed by the mathematician Alfred Haar in
1909. At each level the Haar wavelet change separated a discrete
sign into two parts with half of its length: a high sub-band and
low sub-band. The low sub-band decayed at the first level. One
of the most creative changes that can be utilized to alter a sign
from spatial to recurrence space and the other way around is
wavelet change. Wavelet change, and other related changes,
can be viewed as a second era of changes. Wavelets are character-
ized as motions of short waves that deteriorate quickly [20].
Besides, they have a tremendous number of utilizations that can
be actualized in different fields, for example, signal preparing,
information compacting, unique finger impression checking,
smoothing, picture denoising, and discourse acknowledgment.
It has been noted that wavelet change can be applied to the steg-
anography procedure to expand the limit as well as the power [21].
One of the wavelet change families known as “Haar” has been
actualized in this work. It changes over an image from spatial
domain to frequency domain by applying vertical and horizontal
activity, respectively.

5.1 2D Haar wavelet transform

Time space is conveyed over high-pass and low-pass channels
to evacuate high and low frequencies correspondingly in 2D