Page 31 - Handbook of Deep Learning in Biomedical Engineering Techniques and Applications

Chapter 1 Congruence of deep learning in biomedical engineering 19

HAAR wavelet transform. Discrete wavelet change comprises two
classes of signal: low sub-band and high sub-band. DWT utilizes
two capacity sets: scaling and wavelet, which are associated with
low- and high-pass channels. Deterioration follows the way of
isolating time distinctness. Haar wavelet works on information
that is computed by expansion and taken away from adjoining
components. The wavelet works first on adjoining even segments
and then on the component of neighboring vertical components.
One fundamental element of Haar wavelet change is that the
change is antithetical. Each change registers the energy of the in-
formation repositioned to the upper left-hand corner.
In Fig. 1.8, the detail of images for each resolution can be clas-
sified as diagonal (HH), vertical (LH), and horizontal (HL). The
operations can be repeated on the low (LL) sub-band.
First, scan the pixel from left to right in the horizontal direc-
tion. Then, perform the addition and subtraction operation in
the neighboring pixels. Store the sum on the left and the differ-
ence on the right. Repeat this operation until all the rows are
processed. The letter L represents low frequency and H repre-
sents high frequency.
Second, scan the pixels from top to bottom in the vertical
direction. Perform the addition and subtraction operation on
neighboring pixels and store the sum on the top and the differ-
ence on the bottom. Repeat this operation until all the columns
are processed.
Finally, we will obtain four sub-bands denoted as LL, HL, LH,
and HH, respectively. LL sub-band is low frequency and looks
very similar to the original image.

5.2 Huffman encoding technique
Huffman code for the most part is a method used for informa-
tion compression. Huffman algorithm applies covetous method-
ology that considers the event of each character and conveys an
ideal string of parallel letters. Huffman coding strategies are
utilized for diminishing the measure of bits required to represent
series of images. It is a variable length code that appoints short
length codes to utilized images as often as possible, and long
length codes to the images showing up less often. Huffman codes
are ideal codes that map one image to one code word. For picture
pressure, Huffman coding doles out a parallel code to each pixel
power value and a 2D p q picture is changed to a 1D bit stream
with length not exactly p q. Huffman encoding is applied to
secret objects (pictures/text) and afterward each piece of Huff-
man code of the secret object (picture/text) is installed inside
the spread picture.