1. Transformations
Dr. Amy Zhang

2. Reading
• Hill, Chapter 4, Section 4.5
• Hill, Chapter 5

3. Vectors and Points Recap
Vectors, basis, frames, and points:

4. Geometric Transformations
Functions to map points from one place to another
Geometric transformations can be applied to:
Drawing primitives (lines, triangles)
Pixel coordinates of an image

5. Use of Transformations
Map points from one coordinate system to points in another coordinate system
Change the shape of objects
Position objects in a scene
Create multiple copies of objects in the scene
Projection for virtual cameras
Animations

6. Outline Matrix Algebra * Linear Transformations
Multiple Transformations
OpenGL Geometry Transformations

7. Matrices Matrix: An array of elements that follow certain rules.
We use uppercase letters to indicate a matrix:
A matrix that has n rows and m columns is said to be of order n x m.
We always mention the numbers of rows first.

8. Square Matrices
A matrix is square if the number of row equals the number of columns, i.e., if n = m.
Common square matrices:
Zero matrix: a square matrix with all 0’s.
Identity matrix: all 0’s, except elements along the main diagonal which are 1’s.
The diagonal of a square matrix:

9. Row and Column Vectors A 1 x n matrix is called a row vector.
A n x 1 matrix is called a column vector.
We will use lower case bold letters to indicate vectors.
In graphics (and in this class) we mostly use column vectors.

10. Transpose of a Matrix The transpose of A is written as AT.
AT is formed by interchanging rows and columns of matrix A.
The transpose of a row vector is a column vector.

11. Addition and Subtraction
Addition and subtraction of matrices is defined by addition and subtraction of their corresponding elements.
Note: We can only add / subtract matrices of the same order.

12. Multiplication by a Scalar
To multiply a matrix by a scalar, we multiply each element of the matrix by the number.
Examples:

13. Matrix x Vector Multiplication
If v is a column vector, M goes on the left:
If v is a row vector, MT goes to the right:
We will use column vectors.

14. Matrix x Column Vector
We say the column vector v is pre‐multiplied by M.
The number of rows in M is equal to the number of elements in v.

15. Matrix x Column Vector
The product Mv is a linear combination of the columns:

16. Examples
Compute the matrix‐vector product:

17. Matrix Multiplication
Just apply matrix‐vector multiplication a few times.
Remember: the new vector ( ) is a linear combination of columns.

18. Just apply matrix‐vector multiplication a few times.

19. Just apply matrix‐vector multiplication a few times.

20. Matrix Multiplication
Examples:

21. Properties of Matrix Multiplication
(AB) C = A (BC)
A (B + C) = AB + AC
A (sB) = sAB, where s is a scalar.
(AB)T = BT AT
But:
Example:

22. Outline Matrix Algebra Linear Transformations Multiple Transformations
OpenGL Geometry Transformations
The Viewing Transformation

23. Linear Transformations
A linear transformation L of vectors is just a mapping from v to L(v) that satisfies the following properties:
We can use matrices to express linear transformations of points:

24. Notational Convention
Consider the point P in frame
We will use M to mean “transformed”
The point is transformed with respect to the basis that appears immediately to the left of the transformation

25. Translations
glTranslatef(dx, dy, dz);

26. Translations We have already see how points can be displaced:
For every translation, there exists an inverse function which undoes the translation:
There also exists a special translation, called the identity, that leaves every point unchanged:

27. Groups and Composition
For translations:
There exists an inverse mapping for each function
There exists an identity mapping
When these conditions are met by any class of functions, that class is closed under composition
i.e., any series of translations can be composed to a single translation with matrix multiplication
Mathematically speaking, translations form an algebraic group

28. Rotations Rotation about the origin glRotatef(theta, vx, vy, vz);
theta in degrees, (vx, vy, vz) define axis of rotation

29. Rotations Counter‐clockwise rotation by about the z-axis:
Rotation about the x- and y-axis:

30. Orthonormal Matrices The rotation matrix M has certain properties:
The norm of each row/column is one:
The rows/columns are orthogonal:
We say that rotation matrices are orthonormal.
The inverse of an orthonormal matrix is its transpose

31. Rotation about an Axis
• Rotation about an arbitrary unit vector k = [kx, ky, kz] (Rodrigues Formula):

32. Rigid Body Transformations
The union of translation and rotation functions defines the Euclidean group, also known as rigid body transformations
Properties of rigid body transformations:
They preserve distances
They preserve angles

33. Scaling Scaling about the origin: glScalef( sx, sy, sz);
Each vertex is moved:
sx times farther from the origin in x‐direction
sy times farther from the origin in y‐direction
sz times farther from the origin in z‐direction

34. Uniform Scaling
Uniform scaling by:

35. Reflection about X and Y
Reflection about x-axis:
Reflection about y-axis:
Reflection about x- and y-axis:

36. Similarity Transformations
Add reflections and uniform scaling to the rigid body transformations
Properties of similarity transformations:
Angles are preserved
Distances between points are changed by a fixed ratio
Maintains a “similar” shape (similar triangles, circles map to circles, etc.)

37. Non‐Uniform Scaling
An unbalanced scaling distorts shape:

38. Skews or Shears Along x: Along y:
Shears matrices have a determinant of 1:
The area of the sheared figure stays the same.

39. 3D Shears Shears along different planes:
along y-z plane, along x-z plane, along x-y plane
Shears parallel to different axis:
along x-axis, along y-axis, along z-axis

40. No OpenGL statement: we can load and multiply by arbitrary matrices
glLoadMatrixf(m)
glMultMatrixf(m)
The matrix m is a one dimension array of 16 elements which are the components of the desired 4 x 4 matrix stored by columns
In glMultMatrixf, m multiplies the existing matrix on the right

41. Matrix Stacks
In many situations we want to save transformation matrices for use later
Traversing hierarchical data structures
Avoiding state changes when executing display lists
OpenGL maintains stacks for each type of matrix
Access present type (as set by glMatrixMode) by
glPushMatrix()
glPopMatrix()

42. Affine Transformations
Affine transformations add non-uniform scales and shears to the similarity transformations
Properties of affine transformations:
They preserve our selected plane (sometimes called the affine plane)
They preserve parallel lines
Affine transformations are what we will mostly deal with
Are there other linear transformations?

43. Projective Transformation
The most general linear transformation that can be applied to 2D points
There is something different about this group of Transformations

44. Types of Transformations
Rigid Body (preserve distances)
Translation and rotations
Similarity (preserve angles)
Reflections
Uniform scale
Affine (preserve parallel lines)
Non-uniform scales
Shears
Projective (lines remain lines)
Non-linear (lines become curves)
Twists, bends, warps, morphs, ...

45. Outline Matrix Algebra Linear Transformations Multiple Transformations
OpenGL Geometry Transformations

46. Multiple Transformations
It is quite frequent that we want to apply more than one transformation.
Example:
Translate by vector t.
Rotate by 　degrees about the origin.
If we call these transformations T and R, and the resulting transformation Q, we get:

47. Combined Matrix Expression
Each transformation is expressed in matrix form:
We can compute the combined matrix as follows:
The matrices appear in reverse order of how the transformations are applied

48. Matrix Multiplication
The new, combined matrix is:
Again: This means we are performing the translation with MT first, followed by the rotation with MR
The order matters AB is unequal to BA

49. The Order Matters

50. 2D Rotation about a Point
Translate by t = [‐h, ‐k]
Rotate CCW around origin by 　.
Translate by t’ = [h, k]

51. 2D Rotation about a Point
Writing the composite matrix:

52. Two Points of View Each step is a change of coordinates:
Each step is a change of frames:

53. Outline Appendix: Matrix Algebra Linear Transformations
Multiple Transformations
OpenGL Geometry Transformations

54. Scenes, Actors, Cameras
Use an analogy to classical theatre or photography
Virtual world is called a scene
We call objects in the scene actors
A camera specifies our viewing position and certain viewing parameters (focal length, image size, etc.)
Use 3D affine transformations to position and move actors and cameras in the scene

55. The Scene

56. World, Object, Camera Frames
Use the global world coordinates 　 to place actors and cameras within the scene
Define points (vertices) of objects in some convenient local object coordinates oT　
Define eye / camera coordinates 　for the camera
There could be more than one camera in the scene

57. OpenGL Coordinate Spaces
These coordinate spaces are connected by transformations as follows:

58. Modeling Transformation
The modeling transformation orients / places objects within the world space

59. Viewing Transformation
The viewing transformation maps points from world space into eye space

60. Projection Transformation
The projection transformation maps the viewing frustum to clip space, a cube that extends from -1 to 1 in x, y, and z

61. OpenGL Coordinate Spaces
Eye
World
Screen (3D)
Clip

62. Why does OpenGL do this?
Normalization allows for a single pipeline for both perspective and orthogonal viewing
We stay in four dimensional homogeneous coordinates as long as possible to retain three dimensional information needed for hidden‐surface removal and shading
We simplify clipping

63. Transforming Normals Or:
What happens to normals under affine transformations?
The normal to a surface is a vector that is orthogonal to the tangent plane.
The tangent plane is the plane of vectors that are defined by subtracting nearby surface points.
Or:

64. Transforming Normals
Suppose we transform all points with the affine matrix A.
What vector remains orthogonal to the tangent vector?
Or:

65. Transforming Normals
So the coordinates of the normal are transformed using the inverse transpose of the affine matrix A
If A is a rotation, then the inverse transpose is the same matrix A
If A is a diagonal matrix (non-uniform or uniform scaling), then its inverse transpose is the same as its inverse A-1
In OpenGL, we simply transform the normals with the inverse transpose of the modelview matrix

66. The end
Questions and answers