Title of Seminar/ IT

Chapter-1

INTRODUCTION AND OBJECTIVE

Image segmentation [1] plays an important role in human vision and pattern recognition
fields. Several studies have shown that segmentation based on texture [2] features can
improve the accuracy of the interpretation. A problem that frequently arises when
segmenting an image is that the number of feature variables or dimensionality is often quite
large. It becomes necessary to decrease the number of variables to a manageable size and at
the same time it is useful to retain as much discriminatory information as possible. The other
main difficulty of traditional texture segmentation is the lack of adequate tools to
characterize different scales of textures effectively. Recent developments in multi-resolution
analysis such as wavelet transform help to overcome this difficulty.

For image analysis [1], color and texture are two most important properties, especially
when dealing with real world images like satellite etc. Analysis of image those are not related
to real world or classical images only take into account the pixel grey levels [1], which
represent the total amount of visible light at the pixel’s position. The performance of such
schemes can be improved by adding color information. The color of a pixel is typically
represented with the RGB tri-stimulus color of a pixel is typically represented with the RGB
tri-stimulus values, each corresponding to the Red, Green and Blue frequency bands of the
visible light spectrum. Color is then a feature in the three-dimensional color space like RGB,
contains the information regarding the spectral distribution of light complementary to the
grey-level information. An important topic when processing color images is their
representation. The RGB representation is frequently being transformed into other color
spaces for my work I used Euclidean color space instead of Riemannian space.

1.1 Color Spaces

A digital image forms square picture element called pixels, when displayed on a viewing
device like computer monitor. And that image is a two dimensional matrix of millions of
pixels and each pixel its individual size, address, and color representation.

Session 2021-22 Page 1

A color image contains pixels and each of which holds three values red, green, and blue
levels or based upon color model.

Color space is defined as a model for representing color in terms of intensity values.
Typically, a color space defines as a one to four dimensional space. One of the dimensions is
a color, or a color component. A color dimensional space that is one dimension per pixel
represents the grey-scale space. The following three models are commonly used in color
image retrieval system.

1.1.1 Color Model [3]

The RGB color model is composed of three primary colors Red, Green, and Blue. This color
model defines the system that is used in most color CRT monitors. They are considered the
“additive primaries” since the colors are added together to produce the color that is desired
for the system. Cartesian coordinate system is being used by this color model as shown in
Figure 1.1(a) and (b)

(a) (b)

Fig.1.1. RGB coordinates system (a) and RGB color model (b)

The HSV stands for the Hue, saturation, and Value and these are based on the artists (Tint,
Shade, and Tone). The HSV coordinate system shown in Figure 1.2 (a) and (b) used hex cone
structure. This represents color intensity, which is decoupled from the color information in
the target image. The hue and saturation components of this model are intimately related to
the way human eye perceives color resulting in image processing algorithms with
physiological basis. Hue varies from 0 to 1.0, the corresponding colors vary from red through

Session 2021-22 Page 2

yellow, green, cyan, blue and magenta, moving back to red, so that there are actually red
values both at 0 and 1.0. Saturation varies from 0 to 1.0, the corresponding colors (hues) vary
from unsaturated (grey shades of) to fully saturated on white component. Value or
brightness, varies from 0 to 1.0, the corresponding colors becomes brighter.

Red-Green Sector:

o O
for 0 < H ≤ 120

1 1
b= ( 1−S )= 1+
3 3
ScosH
[
cos ( 60O −H ) ]
, g=1−(r+ b) (1.4)

Green-Blue Sector:

o O
for 120 < H ≤ 240

1 1
r = ( 1−S ) ,¿ 1+
3 3 [
ScosH
cos ( 60 O−H ) ]
,b=1−(r + g) (1.5)

Session 2021-22 Page 3

REFERENCES
[1] "10 Machine Learning Methods that Every Data Scientist Should Know", Medium, 2019.
[Online]. Available: https://towardsdatascience.com/10-machine-learning-methods-that-
every-data-scientist-should-know-3cc96e0eeee9. [Accessed: 20- Jul- 2019].

[2] "Machine learning", En.wikipedia.org, 2019. [Online]. Available:

[3] John C. Russ, “The Image Processing Handbook”, ISBN 0-8493-2516-1, CRC Press,
Taylor & Francis Group, 2011

[4] C. H. Chen, L. F. Pau, P. S. P. Wang (eds.), “The Handbook of Pattern Recognition and
Computer Vision ”, 2nd Edition, by pp. 207-248, World Scientific Publishing Co., 1998.

[5] Rafael C. Gonzalez, Richard E. Woods, “Digital Image Processing”, 3 rd Edition, ISBN
978-81-317-2695-2, Pearson Education, 2009.

[6] Xiuqi Li, Shu-Ching Chen, Mei-Ling Shyu, Borko Furht, “Image Retrieval by Color,
Texture, and Spatial Information”, 2009.

[7] Jain, Anil K., “Fundamentals of Digital Image Processing”, ISBN 0-13-336165-9,Prentice
Hall. pp. 68, 71, 73.

Docs.python.org. 2020. 2. Using The Python Interpreter — Python 3.9.1Rc1 Documentation.

Use http://www.citethisforme.com/ to generate correct format of

Session 2021-22 Page 4