Republic of the Philippines

Laguna State Polytechnic University

ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

LSPU Self-Paced Learning Module (SLM)

Course PERCEPTION AND COMPUTER VISION
Sem/AY First Semester/2021-2022
Module No. 2
Lesson Title SHAPE REPRESENTATION
Week
6-9
Duration
Date November 8-December 3, 2021
Description This lesson will discuss the shape representation. This will give the idea of a good
of the representation. Also, this will various strategies in shape representation.
Lesson

Learning Outcomes
Intended Understand how shape is represented in computer vision.
Learning
Outcomes
Targets/ At the end of the lesson, students should be able to:
Objectives • Define what shape representation is.
• Know the attributes of a good representation.
• Understand different popular strategies in shape representation.

Student Learning Strategies

Online Activities A. Online Discussion via Google Meet

(Synchronous/ You will be directed to attend in a One and a half-Hour class discussion
about the Introduction to Perception and Computer Vision. To have
Asynchronous) access to the Online Discussion, refer to the link provided on our Google
Classroom.

The online discussion will happen on October 11-November 5, 2021,

between 7:00AM-5:00PM based from your class schedule.

(For further instructions, refer to your Google Classroom and see the
schedule of activities for this module)

B. Learning Guide Questions:

1. What is Shape Representation?
2. How shape is represented in a computer vision system?

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Note: The insight that you will post on online discussion forum using Learning Management System
(LMS) will receive additional scores in class participation.
Lecture Guide
Shape is an important visual feature and it is one of the basic features used to
describe image content. Under this lesson, we shall briefly discuss computer vision
shape representation.

Shape Representation

Shape is an important visual feature and it is one of the basic features used to
describe image content. Shape representation generally looks for effective and
perceptually important shape features based on either shape boundary information or
boundary plus interior content

Shape Representation - what for? To reason about an entity we must first

represent the entity.

○ Real world Image processing

○ Shape representation
○ Image understanding/ Shape recognition

Definition
Offline Activities
(e-Learning/Self- Table 1. Definition and its example.
Paced)
Definition Example

Object
- An object is something that can be ball
seen or touched, a material thing.

Abstraction ● ball
- Idea of quality separate of actual ● sphere
examples

Representation 𝑟2 = 𝑥2 + 𝑦2 + 𝑧2
- A way of symbolizing an object

Attributes of a Good Representation

❖ Sufficient
- Is this representation sufficient enough? Depends on the application
❖ Wide domain
- Able to represent many different classes of entities
e.g number of elements in a queue

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

❖ Unique
- every distinct member of its domain has a single distinct representation.
e.g not unique - dog
unique- pitbull / collie / cocker- spaniel

❖ Unambiguous
- An entity may have different representations but no two distinct entities may
have a common representation.
❖ Generative
- Capable of directly generating (recovering) the represented entity
❖ Stable
- small perturbations do not induce large changes in the representation of the
entity.
❖ Convenient
- A representation may exhibit all the characteristics that we have discussed so
far and yet not be convenient for a task.
e. g. an assembly line robot whose task is to filter out rectangles form other shapes
might use chain code representation. Yet for a police computerized camera that is
supposed to compare faces in the crowd to a DB of suspects this is not enough.

Popular Strategies
❖ Volume based :
➢ Describe the object volumetrically
➢ Combination of primitive volumes commonly used, eg. cubes,
tetrahedra, and discs
➢ Provide more access to global relationships
➢ Do not provide direct surface information about the object
➢ Can represent only closed (boundaryless) surfaces
➢ e. g. Symmetric Axis Transform
❖ Surface based :
➢ Surface of object represented by a single closed parametric grid.
➢ Better suited for partial surfaces.
➢ e. g. Parametric Bicubic Patches Gaussian- Image Representations
❖ Others:
➢ Chain code.
➢ Distance vs. Angle.
➢ Fourier Transform Moment
➢ Generalized Cylinders
➢ Visual Potential

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Chain Code

Figure 1. Example of a chain code.

Figure 2. Example of a chain code with 4-neighbor and 8-neighbor.

Figure 3. Chain Code independent of location, starting point and orientation.

Attributes
➔ wide domain
➔ unique
➔ unambiguous
➔ generative - 2 D only stable - depends on tolerance

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Distance vs. Angle

Figure 4. Represent a 2-D object boundary in term of a 1-D function of radial distance with
respect to Ө.

● Find a point of balance.

● Measure distance to edges.
● Plot the graph of distance vs. angle.

Fourier Descriptor
Attributes:
➔ wide domain
➔ unique
➔ unambiguous
➔ generative stable - depends upon tolerance

Figure 5. Shows the Translation, Scale, Rotation, Boundary representation and Fourier
transform.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Moment
• Find center of mass

• The moment:

• since we get invariant values the moments are not affected by transition or
rotation.
• By normalizing the values, we can make the moment be indifferent to scaling
Attributes:
➔ wide domain
➔ unique - no
➔ unambiguous - no
➔ generative - no
➔ stable – no

Parametric Bicubic Patches

By taking a set of four cubic curves with parameter s and using these four
curves to generate a set of four control points for any given value of s, and then using
these control points to define a parametric curve with parameter t we can generate a
surface called a bicubic patch.

• Split the Object to simpler bicubic patches that can be represented by

relatively simple mathematical equations.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Figure 6. Parametric Bicubic Patches.

Piecewise Parametric Surfaces

Surface is partitioned into a parametric surface.

Figure 7. Piecewise Parametric Surface.

Parametric Patches
Point Q(u,v) on the patch is the tensor product of parametric curves defined by
the control points.

Figure 8. Parametric Patches.

Bezier Patches

Figure 9. Bezier Patches.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Properties:
- Interpolates four corner points
- Convex hull
- Local control

Figure 10. Example of Bezier patches.

Attributes
• generative
• stable
• popular in computer graphics
• Possibility of several equally acceptable bicubic approximations to any given
surface makes it inappropriate for surface matching

Symmetric Axis Transform

Figure 11. Symmetric Axis Transform.

● A.k.a Blum Transform, Medial Axis Transform.

● Formally: the object is the logical union of all its maximal discs.

Description :
- Locus of centers a.k.a symmetric axis.
- Radius at each point.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Figure 12. An example showing maximal spheres and symmetry axis.

• May be thought of as grass fire spreading from the border inwards.

Attributes:
➔ wide domain
➔ unique
➔ unambiguous
➔ generative
➔ not stable - small changes affect dramatically
➔ Primarily used in biological applications.

Generalized Cylinders
● Binford 1971
● An extension of SAT.
● Shape represented by an ordinary cylinder sweeping the cross-section along
an arbitrary space curve (axis/spine).
● Decomposition of 3D shape-description problems into lower-order problems

Figure 13. 3D Shape.

• A generalized cylinder is thus defined by

o a cross section,
o an axis
o a sweeping rule.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Figure 14. A Generalized Cylinders.

Attributes:
➔ wide domain - depends on implementation ( redundancy vs. domain
➔ unique
➔ generative
➔ stability is doubtful
➔ decomposition of complex structures is difficult.
➔ used mainly for object recognition
➔ been used in working systems.

Gaussian Image
• Surface normal vector information for any object can be mapped onto a unit
sphere, called the Gaussian sphere.
• Mapping is called the Gaussian image of the object.

Figure 15. Gaussian Image.

Extended Gaussian Image

• The mapping is: Surface normals for each point of the object are placed so that
their
o tails lie at the center of the Gaussian sphere
o heads lie on a point on the sphere appropriate to the particular
surface orientation

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Figure 16. Extended Gaussian image.

We can extend this process so that

• a weight is assigned to each point on the Gaussian sphere equal to the area of
the surface having the given normal
• This mapping is called the extended Gaussian image (EGI).
• Weights are represented by vectors parallel to the surface normals, with length
equal to the weight

Figure 17. Another Example of Extended Gaussian Image.

Attributes
• wide domain
• not unique
• not unambiguous
• not generative
• stable - with limitations
• invariant to translation and scaling
• position independence

Visual Potential
• Aspect - topological structure of the singularities in a single view.
• A graph in which each node represents an aspect of the object, and each edge
the possibility of transiting from one aspect to another under motion of the
observer [Turner74].
• Represents in a concise way any visual experience an observer can obtain by
looking at the object when traversing any orbit through space.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
ISO 9001:2015 Certified
Province of Laguna
Level I Institutionally Accredited

Figure 18. Visual Potential.

Attributes:
➔ wide domain
➔ unique
➔ unambiguous
➔ generative
➔ not stable

Summary
• The above methods are deterministic; in practice, uncertainties in shape that
result from the noise in measurement have to be considered [Ayache88,
Ikeuchi88]
• The above methods are based on geometric models, they are more appropriate
for describing specific objects, particularly artificial ones with regular
structures
• Symbolic models are more appropriate for natural objects that are better
defined in terms of generic characteristics (e.g.. small, red, rough) than precise
shape; useful for matching

Engaging Activity

Research for the following and know its association to computer vision
specifically in shape representation:

• Object
• Abstraction
• Representation

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
Province of Laguna
ISO 9001:2015 Certified
Level I Institutionally Accredited

Performance Tasks
Performance Task 1

Research at least one related article to computer vision, analyze and discuss how they discussed the
shape representation. Please put the reference.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
Province of Laguna
ISO 9001:2015 Certified
Level I Institutionally Accredited

Understanding Directed Assessment

Rubric for Essay
Criteria 1 2 3 4
Quality of Writing • Piece had no • Piece had little • Piece was • Piece was
style or voice style or voice written in an written in an
• Gives no new • Gives some new interesting style extraordinary
information and information but and voice style and voice
very poorly poorly organized • Somewhat • Very informative
organized informative and and well
organized organized
Grammar, Usage & • So many • A number of • Few spelling and • Virtually no
Mechanics spelling, spelling, punctuations spelling,
punctuation and punctuation or errors, minor punctuation or
grammatical grammatical grammatical grammatical
errors that it errors errors errors
interferes with
the meaning
Adopted from: https://www.thoughtco.com/essay-rubric-2081367

Rubric for Research Paper

Criteria 4 3 2 1

Integration The paper The paper The paper The paper does not
of demonstrates that demonstrates that demonstrates that demonstrate that
Knowledge the author fully the author, for the the author, to a the author has fully
understands and has most part, certain extent, understood and
applied concepts understands and has understands and has applied concepts
learned in the course. applied concepts applied concepts learned in the
Concepts are learned in the learned in the course.
integrated into the course. Some of the course.
writer’s own insights. conclusions,
The writer provides however, are not
concluding remarks supported in the
that show analysis body of the paper.
and synthesis of
ideas.

Topic focus The topic is focused The topic is focused The topic is too The topic is not
narrowly enough for but lacks direction. broad for the scope clearly defined.
the scope of this The paper is about a of this assignment.
assignment. A thesis specific topic but the
statement provides writer has not
direction for the established a
paper, either by position.
statement of a
position or
hypothesis.

Depth of In-depth discussion & In-depth discussion The writer has Cursory discussion
discussion elaboration in all & elaboration in omitted pertinent in all the sections of
sections of the paper. most sections of the content or content the paper or brief
paper. runs-on excessively. discussion in only a
Quotations from few sections.

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
Province of Laguna
ISO 9001:2015 Certified
Level I Institutionally Accredited

others outweigh the

writer’s own ideas
excessively.

Cohesiveness Ties together For the most part, Sometimes ties Does not tie together
information from all ties together together information information. Paper
sources. Paper flows information from all from all sources. does not flow and
from one issue to the sources. Paper flows Paper does not flow - appears to be
next without the need with only some disjointedness is created from
for headings. disjointedness. apparent. Author's disparate issues.
Author's writing Author's writing writing does not Headings are
demonstrates an demonstrates an demonstrate an necessary to link
understanding of the understanding of the understanding of the concepts. Writing
relationship among relationship among relationship among does not
material obtained material obtained material obtained demonstrate
from all sources. from all sources. from all sources. understanding any
relationships

Citations Cites all data obtained Cites most data Cites some data Does not cite
from other sources. obtained from other obtained from other sources.
APA citation style is sources. APA citation sources. Citation
used in both text and style is used in both style is either
bibliography. text and inconsistent or
bibliography. incorrect.
Adapted from: Whalen, S. “Rubric from Contemporary Health Issues Research Paper”

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
Province of Laguna
ISO 9001:2015 Certified
Level I Institutionally Accredited

Learning Resources
• Students’ Handbook
• Forsyth, D., & Ponce, J. (2011). Computer vision: A modern approach (p. 792). Prentice hall.
• Sonka, M., Hlavac, V., & Boyle, R. (2007). Image processing, analysis and computer
vision. Thomson, 3ed.
• Szeliski, R. (2010). Computer vision: algorithms and applications. Springer Science & Business
Media.
• https://machinelearningmastery.com/what-is-computer-vision/
• https://www.bharathuniv.ac.in/colleges1/downloads/courseware_ece/notes/BEC007%20%
20-Digital%20image%20processing.pdf
• http://web.mit.edu/manoli/ecimorph/www/code/MMorph.html
• http://www.cs.ubc.ca/~clee/cg/proj/index.html
• http://www.dai.ed.ac.uk/CVonline/LOCAL_COPIES/MARSHALL/node53.html#figuresame_E
GI
• http://www.cs.uwa.edu.au/~cheng/index.html

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION

 Republic of the Philippines
Laguna State Polytechnic University
Province of Laguna
ISO 9001:2015 Certified
Level I Institutionally Accredited

LSPU SELF-PACED LEARNING MODULE: PERCEPTION AND COMPUTER VISION