Lecture-2

By : Dr. Shimaa Saber

1
Course Ethics
 Agenda

Revision on Computer Vision

Revision on Computer Vision System

Data Collection

3
 Revision on Computer Vision
meaning, application

4
 Human Perception
Humans have highly sophisticated skills for sensing
their environment and acting according to their
observations. e.g.

 Detecting and Recognizing a face.

 Understanding spoken words.
 Reading handwriting.
 Distinguishing fresh food from its smell.

We would like to give similar capabilities to

machines
5
 Computer Vision meaning
 Computer vision can be defined as a scientific field that
extracts information from digital images. The type of
information gained from an image can vary from
identification, space measurements for navigation, or
augmented reality applications.

 Computer vision is building algorithms that can understand

the content of images and use it for other applications.

6
 Examples of applications
• Shape and motion capture are new techniques
used in movies like Avatar to animate digital
Special effects
characters by recording the movements played by
a human actor.

3D modeling • Computer vision is used to combine all the photos

into a single 3D model.

• Recognize the location where a photo was taken.

Scene With computer vision, we can do a search on
recognition Google using an image as the query.
• Autonomous driving is one of the hottest
Self-driving applications of computer vision. Companies like
cars Tesla, Google, or General Motors compete to be
7 the first to build a fully autonomous car,
 Examples of applications
Optical • Handwritten: sorting letters by postal code,
Character input device for PDAs.
Recognition • Printed texts: reading machines for blind
(OCR) people, digitalization of text documents.
• Face recognition, verification, retrieval.
Biometrics • Finger prints recognition.
• Speech recognition.
Diagnostic • Medical diagnosis: X-Ray, EKG analysis.
systems • Machine diagnostics, waster detection.
• Automated Target Recognition (ATR).
Military
applications • Image segmentation and analysis (recognition
from aerial or satellite photographs).
8
 Revision on Computer Vision System

9
 Computer Vision System
Real world

Sensor

Pre-processing

Segmentation

Feature Extraction

High-level processing

10 Decision Making
 Computer Vision System
Real world

Sensor

Pre-processing

Segmentation

Feature Extraction

High-level processing

11 Decision Making
 Computer Vision System
Real world
Before a Computer Vision system can be applied
Sensor to image data in order to extract some specific
pattern, it is usually necessary to process the
Pre-processing data in order to assure that it satisfies certain
assumptions implied by the method.
Segmentation

Feature Extraction

High-level processing

12 Decision Making
Noise removing
 Computer Vision System
Real world
Before a Computer Vision system can be applied
Sensor to image data in order to extract some specific
pattern, it is usually necessary to process the
Pre-processing data in order to assure that it satisfies certain
assumptions implied by the method.
Segmentation

Feature Extraction

High-level processing

13 Decision Making
Image enhancement
 Computer Vision System
Real world
Image Segmentation is the process by which
Sensor a digital image is partitioned into various
subgroups (of pixels) called Image Objects,
Pre-processing which can reduce the complexity of the
image, and thus analyzing the image
Segmentation becomes simpler.

Feature Extraction

High-level processing

14 Decision Making
 Computer Vision System
Real world  Compute numeric or symbolic information
from the collected data
Sensor
 features at various levels of complexity are
Pre-processing extracted from the image data

Segmentation

Feature Extraction

High-level processing

15 Decision Making Edges Corners

 Computer Vision System
Real world
 classifying a detected object into different categories.
Sensor

Pre-processing

Segmentation
 comparing and combining two different views
Feature Extraction of the same object.

High-level processing

16 Decision Making
 Computer Vision System
Real world
 Making the final decision required for the
Sensor
application
Pre-processing

Segmentation

Feature Extraction

High-level processing

17 Decision Making
 Data Collection
The first step in the computer vision system

18
 Computer Vision System
Real world

Sensor
image
Pre-processing

Segmentation

Feature Extraction

Decision Making
19
 Data Collection
 Depending on the application we can use different types
of sensors to acquire data:
 Cameras
 Microphones
 GPS sensors, gyroscopes
 heart rate, perspiration sensors, blood pressure sensors

 Take care of:

 Noise levels
 Data acquisition speed
 Amount of collected information
20
 What is an image
 An image is a projection of a 3D scene into a 2D projection
plane.

21
 Digital Image
A digital image is a numeric representation of a two-dimensional image
as a finite set of digital values, called picture elements or pixels.

World Camera Digitizer Digital

Image

22 Camera Image Digitalization

 Camera Model

23
 Digital Image
A digital image is a numeric representation of a two-dimensional image
as a finite set of digital values, called picture elements or pixels.

World Camera Digitizer Digital

Image

Camera

24
 Human eye
 Through our eyes, we can see

25
 Camera
 Camera is computer’s eye that is structurally the same the
eye
– Lens does similarly as our lens and cornea
– Sensor receives the light signals to form images

26
 Camera vs. eye

Differences
– Lens focus
• Camera: lens moves closer/further from the film
• Eye: lens changes shape to focus
– Sensitivity to light
• Camera: A film is designed to be uniformly sensitive to light
• Eye: retina is not; has greater sensitivity in dark
27
 Imaging
Images are 2D projections of real-world scenes

• Images capture two kinds of information:

– Geometric: points, lines, curves, etc.

– Photometric: intensity, color.

28
 Camera Model
 The camera is one of the most essential tools
in computer vision.

 Camera models approximate the Complex 3D-2D

relationships

 Goal of Camera Model

To model basic geometry of projection of 3D points, curves,
and surfaces onto a 2D surface, the view plane or image plane.

29
 Pinhole Camera Model
 This camera system can be designed by placing a barrier with
a small aperture between the 3D object and a photographic
film or sensor.

30
 Pinhole Camera Model

pinhole
3D Object
Virtual image 2D image

Image plane
31
 Camera

32
 Pinhole camera model
 The Internal Camera Matrix

𝑦 𝑌
 =
𝑓 𝑧

𝑌
 𝑦 =𝑓 ∗
𝑧

f = focal length
Y = object
z = distance
between object and
pinhole
33
 Pinhole camera model

𝑌
𝑦 = 𝑓 ∗
𝑧
𝑋
𝑥 = 𝑓 ∗
𝑧

34
The focal length of a lens is the distance between its optical center
and its image plane.
 Pinhole camera model
 In a pinhole camera, the image of height 3 cm is formed at a
distance of 5 cm. If the object is situated at a distance of 10 cm,
the height of the object is:

Image height (y) = 3 cm

focal length (f) = 5 cm
Distance ( z) =10 cm

𝑌
𝑦 =𝑓 ∗
𝑧
𝑌
3 =5 ∗
10
Y = 6 cm
35
 Pinhole camera model
 A tree, which is 200 m away from the pinhole, produces an
image of height 1cm, in a pinhole camera of width 20 cm.
Find the height of the tree.

Object distance from pinhole, (z)=200m

Image distance from the pinhole, (f)=20cm=0.2m
Image height, (y)=1cm=0.01m
𝑌
𝑦 =𝑓 ∗
𝑧
𝑌
0.01 = 0.2 ∗
200
Y = 10 m
36
 Pinhole camera model

What is The effects

of aperture size on
the image?

37
 Lens Camera Model
 In modern cameras, we replace the pinhole with a lens, then
all rays of light that are emitted by some point p are refracted
by the lens such that they converge to a single point P

38
 Lens Camera Model

1 1 1
 Lens equation = +
𝑓 𝑑𝑜 𝑑𝑖
ℎ𝑖 𝑑𝑖
 The magnification m = = −
ℎ𝑜 𝑑𝑜
39
  A 4.00-cm tall light bulb is placed a distance of 45.7 cm from
a double convex lens having a focal length of 15.2 cm.
Determine the image distance and the image size.
 Solution
ho = 4.00 cm do = 45.7 cm f = 15.2 cm
1 1 1 1 1 1
= + = +
𝑓 𝑑𝑜 𝑑𝑖 15.2 45.7 𝑑𝑖
1 1 1 1
= − = 𝑑𝑖 = 22.8 𝑐𝑚
𝑑𝑖 15.2 45.7 0.0439

ℎ𝑖 𝑑𝑖 ℎ𝑖 22.8
= − = −
ℎ𝑜 𝑑𝑜 4 45.7
ℎ𝑖 = − 1.99
40
 Pinhole vs. Lens camera model

41
 Lens Camera Model

42
 Image Digitalization

43
 Digital Image
A digital image is a numeric representation of a two-dimensional image
as a finite set of digital values, called picture elements or pixels.

World Camera Digitizer Digital

Image

Image Digitalization
44
 Image Digitization
 Two stages in the digitization process:
 Sampling
 Quantization

j
f x 1 2 3 4 5

1 100 100 100 100 100

2 100 0 0 0 100

y 3 100 0 0 0 100

4 100 0 0 0 100
i
5 100 100 100 100 100

Continuous Image Digital Image

45
f(x,y) g(i,j)  C
 Image Sampling and Quantization
• Sampling and quantization are the two important
processes used to convert continuous analog image into
digital image.
• Image sampling refers to digitization of spatial
coordinates (along x axis) whereas quantization refers
to digitization of gray level values (amplitude (along y
axis)).

• (Given a continuous image, f(x,y), digitizing the

coordinate values is called sampling and digitizing the
amplitude (intensity) values is called quantization.)
Sampling
 When a continuous scene is imaged on the sensor, the continuous
image is divided into discrete elements - picture elements
(pixels)
Quantization
 Choose number of gray levels (according to number of assigned
bits)
 Divide continuous range of intensity values
 Image Type

Binary Gray Color

49
 Images Types
 Binary image :

each pixel has one of

only two discrete
values: 1 or 0.

A binary image is
stored as a logical
array.
50
 Gray image

 gray Images
 size(X) = [m,n]
 values 0:255 or 0:1
 Images Types

 True color Images :

 Each pixel is specified by three

values ( red, blue, and green)
components of the pixel's color.

 MATLAB store truecolor images as

3 data array .
 size(RGB) = [m,n,c]
 values 0:255 or 0:1
52
 RGB information in separate “channels”

53
 Image Quality

54
 Image Quality
Basically, an image is of good quality if :

 a high resolution
Resolution
 not noisy
Blurring
 not blurred Image
Quality
 good contrast. Contrast

Noise

55
 Image resolution
 Expresses how much detail can be seen in it.
 Depends on the number of pixels used to represent a scene.

𝑯𝒊𝒈𝒉 𝒓𝒆𝒔𝒐𝒍𝒖𝒕𝒊𝒐𝒏 𝑳𝒐𝒘 𝒓𝒆𝒔𝒐𝒍𝒖𝒕𝒊𝒐𝒏

56
 Noisy Image
 Gaussian Noise is caused by random fluctuations in the
signal.
 Gaussian noise occurs during the acquisition process that is
inherent due to the low level of illumination

57
 Noisy Image
 Salt and pepper Noise is randomly scattered white or black
(or both) pixels over the image.
 Occurred due to :
 memory cell failure
 Malfunctioning of camera’s sensor cells
 Errors in image digitizing

58
 Noisy Image
 Periodic Noise
An image affected by periodic noise will look like a repeating
pattern has been added on top of the original image.

59
 Image blurring
 Image blurring is caused by incorrect image-capturing
conditions.
 Out of-focus camera, or relative motion of the camera and
imaged object.

60
 Image Contrast
 Contrast can be simply explained as the difference between
maximum and minimum pixel intensity in an image.

61
 Dataset examples

62
 Example (1): Face Detection Project
Large-scale
CelebFaces
Attributes
(CelebA)
Dataset

202,599 face images,

10,177 identities,
40 attribute

63 http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
 Example (1): Face Detection Project

64
 Example (2): Dogs Classification Project

 Stanford
Dogs
Dataset

categories: 120
images: 20,580

65 http://vision.stanford.edu/aditya86/ImageNetDogs/
 Example (3): Flower Category Project

 Oxford 102
Flower
Dataset

102 flower categories.

Each class consists of
between 40 and 258
images.

66 https://www.kaggle.com/datasets/haseeb85/oxford-102-category-flower-dataset
 Example (4): Flower Category Project

 Fruits 360
dataset

classes : 131
images : 90483

67 https://www.kaggle.com/datasets/moltean/fruits
 Example (5): Flower Category Project

 The
CompCars
dataset

1,716 car models.

68 http://mmlab.ie.cuhk.edu.hk/datasets/comp_cars/
 Other dataset
http://homepages.inf.ed.ac.uk/rbf/CVonli
ne/Imagedbase.htm

69
 Summary
 Computer Vision meaning

 Computer Vision applications

 Computer Vision system

 Data collection “Camera & digitalization”

 Image “quality, types”

 Sample of Datasets

70
71