Transformations and Matrices

2D Transformations
2D Transformations  Transformations are functions
 Matrices are functions representations
 What is transformations?  2 ways  Matrices represent linear transformation
 Changes in orientation , size , and shape are accomplished  Object Transformation  {2x2 Matrices}  {2D Linear Transformation}
with geometrical transformations that alters the coordinate • Alter the coordinates descriptions an object  What are they?
description of object. • Translation, rotation, scaling etc.  changing something to something else via rules
 Why the transformations is needed? • Coordinate system unchanged  mathematics: mapping between values in a range set and
 To manipulate the initially created object and to display the  Coordinate transformation domain set (function/relation)
modified object without having to redraw it. • Produce a different coordinate system  geometric: translate, rotate, scale, shear,…
 Why are they important to graphics?
 moving objects on screen / in space
 mapping from model space to screen space
 specifying parent/child relationships
4
 Basic Transformation : Rotation
Basic Transformations : Translation (1/2) Translation (2/2)

 Translation  In matrix form

x tx 
 Definition : repositioning objects along a straight line path P'  P  T      
y ty 
from one position to another  Translation is rigid-body Transformation that moves
 Let P = x  : original position, P = x  : new position objects without deformation
 y  y 
    • every point on the object is translated by the same amount
 We need translation distance (translation vector)  For straight line
• applying translation distance to each line end points
tx  for x direction
T=   For polygon, curves
ty for y direction
 Then, x'  x  tx
y'  y  ty

Basic Transformations : scaling (1/2)

 Scaling
 Definition : alters the size of an objects
 we need scaling factors

SX 0  : for x value

S  
 0 Sy : for y value
 In matrix form
sx 0  x 
P  S  P    
0 sy 
  y
 Then x ' x  sx y '  y  sy
 if sx  sy : Uniform scaling
sx  sy : differential scaling

scaling (2/2) Inverse 2D - Transformations Homogeneous Co-ordinates (1/3)

 if 0  sx  sy  1 : uniform compression
-1
Translaiton : T(dx,dy)  T(-dx,-dy)  Translation, scaling and rotation are expressed
• move objects to the coordinate origin (non-homogeneously) as:
 if 1  sx  sy : uniform Enlargement Rotation : R -1
(θ )  R(-θ )  translation: P = P + T
• move objects farther from the origin
 Scale: P = S · P
 Fixed point scaling : scaling based on a fixed point ( xf , yf )
• An object is scaled relative to the fixed point by scaling distance
Sclaing -1
: S (sx,sy)  S ( 1sx , 1sy )  Rotate: P = R · P
from each vertex to fixed point  Composition is difficult to express, since translation
x'  xf  (x  xf)sx  x  sx  xf(1  sx)
Mirror Ref : M x-1  M x
not expressed as a matrix multiplication
y'  yf  (y  yf)sy  y  sy  yf(1  sy)  Homogeneous coordinates allow all three to be expr
essed homogeneously, using multiplication by 3 ´ 3
• when (xf,yf)  (0,0) M y-1  M y matrices
 W is 1 for affine transformations in graphics
18 19
 Homogeneous Co-ordinates (2/3) Homogeneous Coordinates (3/3) Classification of Transformations
1. Rigid-body Transformation
 P2d is a projection of Ph onto the w = 1
 Preserves parallelism of lines
plane  For Translation, Rotation, Scaling
 Preserves angle and length
 So an infinite number of points correspon  x   1 0 t x   x   e.g. any sequence of R() and T(dx,dy)
d to : they constitute the whole line  y  0 1 t    y
Translation:
   y   2. Affine Transformation
 s  0 0 1   1
(tx, ty, tw) w  Preserves parallelism of lines
P (x,y,w) h
 Doesn’t preserve angle and length
 x   s x 0 0 x 
w=1 Scaling:  y   0 sy 0   y   e.g. any sequence of R(), S(sx,sy) and T(dx,dy)
  
P2d(x,y,1)  s   0 0 1  1

y Rotation: x  cos - sin 0 x 

 y   sin cos 0   y 
x s  0   
20
   0 1  1
unit cube
22
45 deg rotaton Scale in X not in Y

Properties of rigid-body transformation Concatenation properties.

The following Matrix is Orthogonal if the upper left 2X2 matrix has the
following properties  Matrix multiplication : associative
 r11 r12 tx 
r • ex) For three matrices A, B and C
1. A) Each row are unit vector.
 21 r22 t y 
sqrt(r11* r11 + r12* r12) = 1 A  B  C  (A  B)  C  A  (B  C)
 0 0 1 
B) Each column are unit vector. • Translation or Rotation : additive property commutative
sqrt(c11* c11 + c12* c12) = 1 • scaling : multiplicative property commutative
cos  sin  0
 sin   However, Translation and Rotation : non commutative
 cos 0
2. A) Rows will be perpendicular to each other • order of transformation matrix multiplication is important
 0 0 1
(r11 , r12 ) . ( r21 , r22) = 0
A B  B A
B) Columns will be perpendicular to each other
(c11 , c12 ) . (c21 ,c22) = 0
e.g. Rotation matrix is orthogonal

• Orthogonal Transformation  Rigid-Body Transformation

• For any orthogonal matrix B  B-1 = BT 23

Commutative of Transformation Matrices Composite Transformation (2/2)

Composite Transformation (1/2)
• In general matrix multiplication is not commutative
• For the following special cases commutativity holds i.e. M1.M2 = M2.M1
 Translation  Rotation
 If two successive translation factor (tx1, ty1) and (tx2, ty2) are  two successive rotation  , 
M1 M2 applied to a coordinate point P P  R( 2)  {R( 1)  P}  {R( 2)  R( 1)}  P
Translate Translate • then R( 1)  R( 2)  R( 1   2)
Scale Scale P'  T2(tX2,ty2)  {T1(tx1,ty1)  P}  {T2(tX2,ty2)  T1(tx1,ty1)}  P
P'  R( 1   2)  P
Rotate Rotate • ex) 1 0 tx2  1 0 tx1 1 0 tx1  tx2   two successive rotations are also additive
0 1 ty2   0 1 ty1  0 1 ty1  ty2 
Uniform Scale Rotate   Scaling
0 0 1  0 0 1  0 0 1 
i.e, sx2 0 0 sx1 0 0 sx1  sx2 0 0
• Some non-commutative T(tx2,ty2)  T(tx1,ty1)  T(tx1  tx2,ty1  ty2)  0 0  0  0
 sy2  0 sy1  0 sy1  sy2 
Compositions: 
 0 0 1
0 0 1
 
 0 0 1

 Non-uniform scale, Rotate Original
 Two successive Translation are additive
Transitional S(SX2,SY2)  S(SX1,SY1)  S(SX1  SX2,SY1  SY2)
 Translate, Scale Final
 Rotate, Translate
26
 successive scalings are multiplicative