Page 169 - Algorithm Collections for Digital Signal Processing Applications using MATLAB

158 Chapter 4

Vector 2 = [A B C D E F] (say)

Note that the number of elements shown in the vector is considered as 6.
[Actually the size of the vector is around 200]

Low level feature (LLF) vector after crossover

Vector 1’= [a b c d E F]
Vector 2’ = [A B C D e f]

6. Thus 8 Low level feature vectors for the second generation are obtained
as mentioned below.
LLF1, LLF2, LLF3, LLF4, LLF5, LLF6, LLF7 and LLF8 (say)

7. Compute the Euclidean distance between the feature vector LLF1 with
all the low level feature vectors in the database. Retrieve the image
corresponding to the LLF vector in the database whose Euclidean
distance is smallest. Repeat the procedure for the feature vectors
LLF2, LLF3, … LLF8

8. Thus the image selected for the next iteration is listed below.

9, 4, 16, 3, 13, 18, 45, 43 (say). Note that 9 in the selected list indicates

th
9 image in the database.

9. Repeat the step from 1 to 8 until the user is satisfied with the images
obtained in the latest iteration.
6.2 Example

Consider the Image database consists of 94 Natural sceneries. Every image
is truncated to the standard size 200x200. The feature vector for the
particular image is computed as described below.
The image is divided into sub blocks of size 50x50 with overlapping size
of 8x8.The first feature namely variance of the histogram of the hue content,
are computed for the all the overlapping sub blocks of the image. This is
treated as first part of the feature vector corresponding to the image.
Similarly other features are computed for all the overlapping sub blocks of
the image to obtain other parts of the feature vector. All the parts belonging
to the individual features are combined to obtain the feature vector
corresponding to that particular image.
Thus feature vectors are computed for all the images in the data base.