Module II(8 Hrs) Review of matrix theory, row and column ordering- Toeplitz, Circulant and block matrix, 2D transforms - DFT, its properties, Walsh transform, Hadamard transform, Haar transform, DCT, KL transform and Singular Value Decomposition. Matrix theory Vectors and matrices: - Both 1 d and 2 D can be represented as vectors and matrices - U(n) = u(1) u(2) . . - The nth element of the vector u is denoted by u(n) - A column vector of size N is also called an Nx1 vector N U T K N I . S OTE Downloaded from Ktunotes.in - a matrix A of size MxN has M rows and N columns. N U T K N I . S ..... TaE (1, N ) O a (1,2) a(1,1) A = a(m,n)= a(2,1) a(2,2) ...... a (2, N ) .. ... ... ... a ( M , 1 ) a ( M , 2 ) ..... a ( M , N ) Downloaded from Ktunotes.in Row and column ordering: - xT T x(1,1) x(1,2).....x(1, N ) x(2,1)....x(2, N ).....x( M ,1)....x( M , N ) -for 1 to 1 mapping this is called as a lexicographic or dictionary ordering N I . S E - X is row vector got by stacking each row one after another T O N U T K - Column ordered vector is got by stacking column by column. T xT x(1,1) x(2,1).....x( M ,1) x(1,2)....x( M ,2)....x(1, M ).....x( M , N ) Downloaded from Ktunotes.in T Transposition and conjugate rules: - T * A A *T N I . S T T T OTE AB KBTUAN A A T 1 1 T AB * AB * * Downloaded from Ktunotes.in Toeplitz and circulant matrix - A Toeplitz matrix T is a matrix that has constant elements along the diagonal and the sub diagonals. - A Toeplitz matrix is an n xn matrix Tn = [tk,j; k, j = 0, 1, . . . , n − 1]where tk,j = tk−j , i.e., a matrix of the form N I . S OTE N U T Kt0 t1 t1 t2 t n 1 t0 t1 t 2 ...............t ( n 1) t 1 t0 t0 Downloaded from Ktunotes.in - circulant matrix is a special kind of Toeplitz matrix where each row vector is rotated one element to the right relative to the preceding row vector. - Each of its rows is a circular shift of the previous row N U T K N I . S OTE Downloaded from Ktunotes.in Show each of the following: a. A circulant matrix is Toeplitz but the converse is not true b. The product of 2 circulant matrices is a circulant matrix c. The product of 2 Toeplitz matrices need not be Toeplitz. N U T K N I . S OTE Downloaded from Ktunotes.in Orthognal and unitary matrix: An orthognal matrix is such that its inverse is equal to its transpose A is orthognal if A-1 =A T N U T K N I . S OTE Or ATA=AAT=I Downloaded from Ktunotes.in • A matrix is called unitary matrix if ts inverse is equal to its conjugate transpose A-1= A*T OR AA*T-A*TA=I N U T K N I . S OTE A real orthognal matrix is also unitary but a unitary matrix need not be orthognal Downloaded from Ktunotes.in Problem 1. consider matrices below and check for orthognality and unitary 1 1 1 A [ ]IN S 1. 2 1OT E N U T K2 j B [ ] j 2 j 1 1 c [ ] 1 2 j Downloaded from Ktunotes.in Block matrix - A matrix whose elements are matrices are called block matrix P11 P P21 1 1 P11 1 1 P12 P22 N I . S OTE mxn N U T K P 2 12 pxq 2 2 2 3 3 P 21 3 3 4 4 P 22 4 4 Downloaded from Ktunotes.in - Therefore P is a mxn block matrix with basic dimension pxq. - If block structure is toeplitz or circulant it is called a block toeplitz or block circulant. - if each bolck itself is toeplitz or circulant then we have the doubly block toeplitz or doubly block circulant. Kronecker Product: N U T K N I . S OTE Downloaded from Ktunotes.in Problem A= 1 1 1 -1 B= 1 3 N I . 3 4 S E T O N U T K product of A and B as well Find kronecker as B and A , Are they equal Ans: They are not equal .A kronecker B is not equal to B kronecker A Downloaded from Ktunotes.in Image Transforms Image Transforms are used for image processing and image analysis. Transform is a mathematical tool for moving from one domain to another for performing task in a easy manner. Image transforms are useful for fast computation of convolution and correlation. By using transforms the signals are represented as a set of basis function N U T K N I . S OTE Downloaded from Ktunotes.in Need for transforms Mathematical convenience: convolution in time domain is multiplication in frequency domain To Extract more information: N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in • What does image transform do – It represents a given image as a series summation of a set of unitary matrices N I – A matrix A is a unitary matrix if . A A S E T O N KTU 1 1 *T A A *T – A* is the conjugate of A Downloaded from Ktunotes.in Walsh transform • The basic kernel function of 1D walsh transform is given by 1 N 1 g (m, k ) ( ) (1) bi ( m )bn 1i ( k ) N i 0 IN . S E NOT • WT[x(m)] = KTU 1 X (K ) N 1 x(m) g (m, k ) N m 0 • n represents no of bits to represent a number and bi(n) rep the i th bit of the binary value(from LSB) and n=log 2N Downloaded from Ktunotes.in • Inverse walsh transform N 1 WT-1[X(K)] =x(m)= X ( K )h(m, k ) K 0 2D walsh Transform N I . S OTE N U T K N 1 N 1 1 X ( K , L) N x(m, n) g (m, k , n, l ) n 0 m 0 1 N 1 bi ( m ) bn 1 i ( k ) g (m, k , n, l ) ( ) (1) N i 0 Downloaded from Ktunotes.in bi ( n ) bn 1 i ( l ) Inverse 2D walsh Transform N 1 N 1 x(m, n) X ( K )Eh(Sm.I,Nn, k , l ) T O N K 0K LT 0U N 1 h(m, n, k , l ) (1) bi ( m ) bn 1 i ( k ) bi ( n ) bn 1 i ( l ) i 0 Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in Hadamard matrix • Hadamard matrices seem such simple matrix structures: they are square, have entries +1 or −1 and have orthogonal row vectors and orthogonal column vectors. • A square matrix with elements ±1 and size h, whose distinct row vectors are mutually orthogonal, is referred to as an Hadamard matrix of order h. N U T K N I . S OTE Downloaded from Ktunotes.in Hadamard Matrices • The order N=2 hadamard matrix is given by 1 1 H2 1 1 N I . S OTE N U T K matrix of order 2N • The Hadamard can be generated by kronecker product operation HN HN H 2N HN HN Downloaded from Ktunotes.in • Substituting N=2 H 2 H 2 H4 H 2 H 2 N U T K N I . S OTE Downloaded from Ktunotes.in Haar Transform N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in • Develop hadamard for f(x)={1,2,0,3}------1D • HT for 1D : F=H.f = 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 N U T K = 1 2 0 N I . S OT E 3 6 4 0 2 Downloaded from Ktunotes.in • Develop Hadamard for 2D 2 1 2 1 1 2 1 2 3 2 N I 3 4 3 . S E T O N U 2 3 2 T K • 2D HT : F=H.f.HT =H.f.H Ans: 2 6 6 2 2 2 6 2 2 2 6 2 2 2 34 2 Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in Discrete Cosine Transform N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in Karhunen –Loeve Transform This transform also called as Hotelling Transform. KL transform is a reversible linear transform hat exploits statistical properties of a vector representation The basic functions of KL transforms are orthogonal eigen vectors of the covariance matrix of a data set After a KL transform , most of the energy of the transform coefficients are concentrated within the first few components. Drawbacks of KL transform: It is input dependent and the basic function has to be calculated for each signal model in which it is operated. KL transform require O(m2) multiply / add operations .The DFT and DCT require O(log2m) multiplications. N U T K N I . S OTE Downloaded from Ktunotes.in Applications of KL transform: Clustering Analysis: to determine the new coordinates for the sample data where largest covariance lies on the first axis and next largest on the next axis and so on, and therefore used for dimensionality reduction. Image compression N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in Singular value decomposition(SVD) N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in N U T K N I . S OTE Downloaded from Ktunotes.in