Page 32 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

20 Chapter 1 Congruence of deep learning in biomedical engineering

5.3 Embedding algorithm
Sources of info: secret data (D), cover image (C)
Output: stego image (S)
1. Apply an autoencoder to the input image, which helps to train
as well as compress the image, and then apply the Huffman
encoding procedure in the secret image.
2. Disintegrate the spread picture into four noncovering sub-
groups. These are LL (approximation coefficients), LH (vertical
subtleties), HL (horizontal subtleties), and HH (diagonal
subtleties).
3. The division of the planes is finished by utilizing Haar
channels.
4. Data contained in the LL subgroups of secret pictures is inde-
pendently implanted into various groups of the spread picture.
5. In the wake of implanting the secret picture bit into the spread
picture, backward change is performed to recover them. At
that point it is joined to produce the last stego picture.
Extraction algorithm:
Input: stego image (S)
Output: secret data (D)
1. Apply IWT in stego picture S.
2. The stego picture is isolated into noncovering subgroups. The
subgroups are LL, LH, HL, and HH.
3. Apply the Huffman decoding method in the secret pictures
and the secret pictures are extracted from the corresponding
embedded frequency bands of the cover image.

6. Conclusion and future work

The procedure of stego image embedding as well as the
extraction procedure is described in Fig. 1.11 by using an autoen-
coder, which is a deep learning technique. We can then also use
other deep learning models like CNN, GAN, etc. By using these
types of image encryption as well as decryption methods, hospi-
tals/physicians can preserve patient’s information in the patient
report, MRI, and CT scan data. Steganography is a procedure
that is used for subtly composing messages that may be recovered
from the sender and collector. In this chapter, investigation of
DWT and IWT techniques was effectively actualized and the
results were conveyed. We studied different steganographic
strategies where we used steganographic methods with IWT space
to build the implanting limit. We also correlated the DWT method
with the IWT procedure, which will enable us in the not so distant
future to execute an IWT strategy and produce a key for improving
the concealing limit and achieving better results.