• Objectives: This module will enable the students

Module 1- to:
• Understand Image Formation &
Introduction Representation
• Understand the need for Image Filtering &

to Transformation techniques

Computer
Vision (CV)
 Image Formation
Topics to be
covered Linear Filtering

Color Models
 Projection of 3D scene into 2D
image

Pin-hole projection
Image
Formation Significance of Lens in Image
formation

Image Transformations
 Pixel Transformations

Histograms
Linear
Filtering Point Operators

Convolution
 • Representation of colors as set of
numbers

Color • RGB
• CMY

Models • HIS (Hue, Saturation &

Intensity)
• YIQ (Luminance, In-phase &
Quadrature)
Image Formation
 A Brief History of Images 1544

Camera Obscura, Gemma Frisius, 1544

Source: Computer Vision Lecture notes, Carnegie Mellon University

Image Formation & Optics
Relationship between P0 & Pi
 Pinhole and the Perspective Projection

Is an image being formed

(x,y) on the screen?
screen

YES! But, not a “clear” one.

scene

image plane ! = # #" "" ! !

y
effective focal length, f’ z
optical
axis pinhole
x

! " = $ #" # " " # ! " !

!! ! $! $ "! "
= = =
"! ! #! ! #! !

Source: Computer Vision Lecture notes, Carnegie Mellon University

 Problems with Pinholes

• Pinhole size (aperture) must be

“very small” to obtain a clear image.

• However, as pinhole size is made smaller,

less light is received by image plane.

• If pinhole is comparable to wavelength !

of incoming light, DIFFRACTION blurs
the image!

• Sharpest image is obtained when:

pinhole diameter " =" ! !!

Example: If f’ = 50mm,
= 600nm (red),
d = 0.36mm

Source: Computer Vision Lecture notes, Carnegie Mellon University

Image Formation using Lens
 Aperture, F-Number

• Aperture : Diameter D of the lens that is exposed to light.

• F-Number (f/#):

Copyright: © Jared C. Benedict.

• For example, if f is 16 times the pupil diameter, then f/#=f/16.

• The greater the f/#, the less light per unit area reaches the image plane.

• f-stops represent a convenient sequence of f/# in a geometric progression.

 Image Formation using Lenses
• Lenses are used to avoid problems with pinholes.
• Ideal Lens: Same projection as pinhole but gathers more light!

i o
Po

Pi

! ! !
• Gaussian Thin Lens Formula: + =
# " !
• f is the focal length of the lens – determines the lens’s ability to refract light
• f different from the effective focal length f’ discussed before!
Image Magnification &Perspective Projection
 Lens Defocus

Source: [Link]
Hand notes
• Lens is used to gather more light
• Lens need to be focused to get a clear image
• For each object plane there will be only one image plane where we
get clear image.
Lens Defocus
Image Formation contd..
Factors Lighting conditions
affecting the
mapping Scene Geometry
from 3D Surface properties
scene to 2D
image Camera Optics
 Simple Image Transformations:

• Translation
• Scaling
• Rotation
Image • Reversal
Transformation
(Geometric Affine Transformation
Transformation)
Projective Transformation

• Perspective projection is an example for

Projective Transformation
 2D Transformations

Reference: [Link]
How Rotation is performed using 3x3 affine
matrix?
• Parameters a,b,c & d are used as shown below to Rotate the given
image in clockwise and counter-clockwise directions.
 3x3 parameters for 4 different Transformations

How in the case of affine transformation, x & y coordinates of an original image f(x,y) gets transformed into new
coordinates?
2D Transformations
Ref: [Link]
 Rigid Transformations preserve the
distances and angles between two
points

Euclidean geometry deals with the

concepts of congruence (when two
geometric figures are identical in shape
and size)

Ref: [Link]
Refence: [Link]
Rotation using Python
Scaling operation using Python
Scaling in both directions with axis adjustment
Translation using Python
Affine Transformations
• Combines Linear Transformations & Translation (Non-linear as origin changes)
• Properties:
• Origin doesn’t necessary map to origin (Translation changes the origin)
• Lines map to lines (Rotation,Shearing,Translation & Scaling will map line to line)
• Parallel lines will remain parallel (All the 4 operations mentioned above will preserve
parallelism)
• Ratios are preserved
• 6 parameters (Degree Of Freedom) is 6.
 Projective Transformations
• Origin doesn’t necessary map to origin
• Lines map to lines
• Parallel lines may not remain parallel
• Ratios are not preserved
• 8 parameters (Degree Of Freedom) is 8.
Hue Saturation & Value (HSV) colour representation
• HSV (Hue, Saturation, Value) color representation offers several advantages compared to RGB (Red, Green,
Blue) when it comes to certain image processing tasks and human perception of color.

• Simplifies Color Manipulation: HSV separates the color information (Hue) from the intensity information
(Saturation and Value), making it easier to manipulate and adjust colors independently. This can be useful
for tasks like color correction and image enhancement.

• Intuitive Hue Control: In HSV, the Hue component represents the dominant color in an image, making it
intuitive to adjust the overall color balance or change the color of an object without affecting its brightness
or saturation.

• Brightness Information Separation: The Value (V) component of HSV represents the brightness or
luminance of the color, which is helpful for tasks that involve adjusting the brightness or contrast of an
image.

• Color Detection and Segmentation: HSV is often used for color-based object detection and segmentation
tasks. It is easier to define a range of hues and saturations to isolate specific colors in an image.