Scribd
Upload
39 views

Lect11 - Image Transforms

Here are the key steps to show the image can be constructed using the given basis images: 1) The basis images form an orthogonal set since they are orthogonal to each other 2) Take the inner product of the original image with each basis image to get the transform coefficients 3) Reconstruct the image by taking the linear combination of the basis images weighted by the transform coefficients 4) Show that the reconstructed image is identical to the original image, proving it can be represented using the given basis images.

Uploaded by

nnehasingh
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
39 views

Lect11 - Image Transforms

Here are the key steps to show the image can be constructed using the given basis images: 1) The basis images form an orthogonal set since they are orthogonal to each other 2) Take the inner product of the original image with each basis image to get the transform coefficients 3) Reconstruct the image by taking the linear combination of the basis images weighted by the transform coefficients 4) Show that the reconstructed image is identical to the original image, proving it can be represented using the given basis images.

Uploaded by

nnehasingh
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 14

Digital Imag

e
Processing
Lecture 11
IMAGE TRANSFORMS

0 Applying Transform we get another image of same size.


0 After applying Inverse Transform we get the original image back.
0 Then what is the use of Image Transformation?
Need for Image Transforms
0 Extract information: feature, low and high spatial frequency content

Spectrum
Frequency domain
TRANSFORM
Signal
Time domain

0 Mathematical convenience
0 Convolution in time <==> Multiplication in frequency
domain (complex) domain (simple)
0 Compact form for data storage
Spatial Frequencies in Image
Processing
0 ‘Spatial frequency’ is used to describe the rate of change of
pixel intensity of an image in space. It can be visualized
using the line profile of the image in any row or column.
Introduction
0 What does the Image Transform do?
It represents the given image as a series summation of a set
of Unitary Matrices.
Categories of image transforms: orthogonal (sinusoidal) and
non-sinusoidal.
0 What is a Unitary Matrix?
A Matrix ‘A’ is a unitary matrix if:
A-1 = A*T where A* is conjugate of A
If a Unitary Matrix is real it is known as ORTHOGONAL
Matrix.
Unitary Matrix --------> Basis Functions
Different Image Transforms

IMAGE TRANSFORMS

BASIS FUNCTION DEPENDING


ORTHOGONAL SINUSOIDAL ORTHOGONAL NON- DIRECTIONAL
ON STATISTICS OF INPUT
BASIS FUNCTION SINUSOIDAL BASIS FUNCTION TRANSFORMATION
SIGNAL

- FOURIER TRANSFORM - HAAR TRANSFORM - HOUGH TRANSFORM


- DISCRETE COSINE - WALSH TRANSFORM - KL TRANSFORM - RADON TRANSFORM
TRANSFORM
- HADAMARD TRANSFORM - SVD TRANSFORM - RIDGELET TRANSFORM
- DISCRETE SINE TRANSFORM
- SLANT TRANSFORM - CONTOURLET TRANSFORM
Basis for a transform
0 Let
  u(n) be a vector of size N,
u(n)=[U(0) U(1) U(2)….. U(N-1) ]
0 Let A be the square transformation matrix (of size N x N)
applied on u(n) to generate resultant v(n) of size N.

Individual element of v is given as


for all 0 ≤ k ≤ N-1
Since A is UNITARY , A-1= A*T. So, to recover u, we need
v=A-1u = A*Tu
for all 0 ≤ k ≤ N-1
Introduction
0 Expanding matrix a(k, n), it is of the form,
 a(1,1) a(1, 2) ... a(1, n) 
 a(2,1) a(2, 2) ... a(2, n) 
A(k , n)   
   
 
 a ( k ,1) a ( k , 2)  a ( k , n ) 
0 a*(k, n) is the nth column vector of matrix A*T
0 These column vectors are usually called the basis vectors of A.
N
u ( n )   a * ( k , n )v ( k )
k 1
n = 0, 1, ……..N-1

Where, u(n) is series summation of basis vectors


For 2-D images
 The forward and transformation is written as:

G(k,l)=T(m,n,k,l) f(x,y)

IMP: to get one value we need to operate on the entire image.


Introduction
      
 
     
0 The kernel matrix 
   
 
        kXl

0 Kronekar product:
Interpretation
0 Any N X 1 sequence can be viewed as vector from origin to
any point in N- dimensional space. And a Unitary transform
can be viewed as a coordinate transformation, rotating the
vector in N-space without changing its length.
0 Thus, a vector of N transform coefficients is generated each
of which is the inner product of the input vector and the
basis vector.
0 Forward transformation = analysis => signal vector is
broken into elemental components (basis vectors). The
transform coefficients specify how much of each component
is found to be present in the mixture.
0 Inverse transformation = synthesis, i.e. reassembling the
original vector from its components.
Questions
0QUESTION
  1
Show that A is orthogonal, where

0 SOLUTION: orthogonal matrices satisfy the relation


A-1 = AT
So, A A-1 = AAT =I
Question 2
0  Prove that the Unitary transform Kernel  1 1 1
(
√2 1 −1 )
Works for the given image .
SOLUTION: g = A x f x AT
f = A*T x g x AT
Question 3
0  Show that for the given image , the image can be
constructed using the following basis images:

SOLUTION: t1 =; t2 =; t3 =; t4 =
Synthesis: t1H1 + t2H2 + t3H3 + t4H4

You might also like