Page 168 - Algorithm Collections for Digital Signal Processing Applications using MATLAB

4. Selected Applications 157

rank if user likes that image and expecting the similar images in the next
iteration.
3. 4 Images are selected from the 8 images using Roulette wheel selection
(Refer chapter 1) as described below.

• Let the rank values assigned for the eight images are represented as the
vector as given below.

[10 90 10 10 10 90 10 90]

• Normalized vector is computed as

[0.0313 0.2813 0.0313 0.0313 0.03133 0.2813 0.0313 0.2813]

• Simulate Roulette wheel

Cumulative summation of the above-normalized vector is displayed below.
[0.0313 0.3125 0.3438 0.3750 0.4063 0.6875 0.7188 1]

Random number is generated as ‘r’. If r is in between ‘a’ and ‘b’ in the
cumulative summation vector song corresponding to the value ‘b’ is selected
as mentioned below.

Generated random number = 0.9501(say)
th
Therefore select 8 image.

Generated number =0.2311(say)
nd
Therefore select 2 image.

Similarly.
th
0.6068-----------------------------------6 image
st
0.0185-----------------------------------1 image.

4. Thus the selected images are 8,2,6,5

5. Low level features corresponding to the image number 8,2 6 and 5 are
computed. The vectors thus obtained are randomly chosen and is
subjected to cross over and mutation operation as described below.

Low level feature (LLF) vector before and after crossover

Vector 1 = [a b c d e f]