*XII. Wavelet Transform Main References R. C. Gonzalez and R. E. Woods, Digital Image Processing, Chap. 7, 2nd edition, Prentice Hall, New Jersey, 2002. ()
 S. Mallat, A Wavelet Tour of Signal Processing, Academic Press, 3rd edition, 2009. () ( Chapter 4 )
*Other References I. Daubechies, Orthonormal bases of compactly supported wavelets, Comm. Pure Appl. Math., vol. 4, pp. 909-996, Nov. 1988. S. Mallat, Multiresolution approximations and wavelet orthonormal bases of L2(R), Trans. Amer. Math. Soc., vol. 315, pp. 69-87, Sept. 1989.  C. Heil and D. Walnut, Continuous and discrete wavelet transforms, SIAM Rev., vol. 31, pp. 628-666, 1989.  I. Daubechies, The wavelet transform, time-frequency localization and signal analysis, IEEE Trans. Information Theory, pp. 961-1005, Sept. 1990. R. K. Young, Wavelet Theory and Its Applications, Kluwer Academic Pub., Boston, 1995. S. Qian and D. Chen, Joint Time-Frequency Analysis: Methods and Applications, Chapter 4, Prentice-Hall, New Jersey, 1996. L. Debnath, Wavelet Transforms and Time-Frequency Signal Analysis, Birkhuser, Boston, 2001.  B. E. Usevitch, A Tutorial on Modern Lossy Wavelet Image Compression: Foundations of JPEG 2000, IEEE Signal Processing Magazine, vol. 18, pp. 22-35, Sept. 2001.
*(1) Conventional method for signal analysis Fourier transform : Cosine and Sine transforms: if x(t) is even and oddOrthogonal Polynomial Expansion (2) Time frequency analysis STFTTime frequency analysis
* time-variant spectrum signal representation12-A Haar Transform8-point Haar transform
*N = 8N = 2N = 4
*General way to generate the Haar transform:where means the Kronecker product
*N = 2k H 0 row
2p + q row hp,q[n]N 1hp,q[n] = 1 when q2kp n < (q+1/2)2kp hp,q[n] = 1 when (q+1/2)2kp n < (q+1)2kp p = 0, 1, , k1, q = 0, 1, , 2p1 k = log2N
* Inverse 2k-point Haar Transform D[m, n] = 0 if m nD[0, 0] = 2k, D[1, 1] = 2k, D[n, n] = 2k+p if 2p n < 2p+1When k = 3,
*(1) No multiplications (2) Input Output (3) ( 1) ( 1 1)(4) localized feature(5) Very fast, but not accurateExample:12-B Characteristics of Haar Transform
*References A. Haar, Zur theorie der orthogonalen funktionensysteme , Math. Annal., vol. 69, pp. 331-371, 1910. H. F. Harmuth, Transmission of Information by Orthogonal Functions, Springer-Verlag, New York, 1972.The Haar Transform is closely related to the Wavelet transform (especially the discrete wavelet transform).
TransformsRunning Timeterms required for NRMSE < 105DFT9.5 sec43Haar Transform0.3 sec128
*12-C History of the Wavelet Transform 1910, Haar families. 1981, Morlet, wavelet concept. 1984, Morlet and Grossman, ''wavelet''. 1985, Meyer, ''orthogonal wavelet''. 1987, International conference in France. 1988, Mallat and Meyer, multiresolution. 1988, Daubechies, compact support orthogonal wavelet. 1989, Mallat, fast wavelet transform. 1990s, Discrete wavelet transforms 1999, Directional wavelet transform 2000, JPEG 2000
*Wavelet continuous / discrete 3 Type 1 Input Output NameContinuous Continuous Type 2 Continuous Discrete Type 3 Discrete Discrete Continuous Wavelet Transform Discrete Wavelet Transform 12-D Three Types of Wavelets discrete wavelet transformcontinuous wavelet transform with discrete coefficients
*Definition: x(t): input, (t): mother wavelet a: location, b: scaling a is any real number, b is any positive real number
Compare with time-frequency analysis: location + modulation
Gabor Transform b12-E Continuous Wavelet Transform (WT)
t-axisf-axisa-axisb-axisGabor Wavelet transform frequency t-axis
a: location, b: scaling The resolution of the wavelet transform is invariant along a (location-axis) but variant along b (scaling axis). If x(t) = (t t1) + (t t2) + exp(j2 f1t) + exp(j2 f2t), f2 b2 f1 b1
t1 t2 t1 t2 STFT WT b-axis
*There are many ways to choose the mother wavelet. For example, Haar basis Mexican hat function 12-F Mother WaveletIn fact, the Mexican hat function is the 2nd order derivation of the Gaussian function.
*(1) Compact Support
support: the region where a function is not equal to zerocompact support: the width of the support is not infinite
Constraints for the mother wavelet:(2) Real(3) Even Symmetric or Odd Symmetric ab
*kth moment:If m0 = m1 = m2 = .. = mp1 = 0, we say (t) has p vanishing moments.Vanish moment 2006(4) Vanishing MomentsQuestion ?
*Vanish moment = 1[Ref] S. Mallat, A Wavelet Tour of Signal Processing, 2nd Ed., Academic Press, San Diego, 1999. the 1st order derivation of the Gaussian function ab
*Vanish moment = 2the 2nd order derivation of the Gaussian function ab
*Similarly, whenthe vanish moment is p
*(5) Admissibility Criterion[Ref] A. Grossman and J. Morlet, Decomposition of hardy functions into square integrable wavelets of constant shape, SIAM J. Appl. Math., vol. 15, pp. 723-736, 1984.
*12-G Inverse Wavelet Transform where(Proof): Sinceif then
*whereIf (t) is real, (f) = *(f), (bf) (bf) = *(bf) (bf) = |(bf)|2 (f1 = bf, df1 = fdb)Therefore, y(t) = x(t).
*12-H Scaling Function scaling function wherefor f > 0, (f) = *(f) (t) is usually a lowpass filter (Why?)
* Wavelet transform a is any real number, 0 < b < b0 reconstruction:(1)(2) b0
*(Proof): If(from the similar process on pages 359 and 360)
*Therefore, if y(t) = y1(t) + y2(t),
y(t) = x(t)
*12-I Property(1) real input real output(2) If x(t) Xw(a, b), then x(t ) Xw(a , b), (3) If x(t) Xw(a, b), then x(t /) (4) Parsevals Theory: When (t) = C1(t),
*12-J ScalogramScalogram Wavelet transform Scalogram
*Continuous WT is good in mathematics.
In practical, the discrete WT and the continuous WT with discrete coefficients are more useful.Problems of the continuous WT(1) hard to implement(2) hard to find (t)12-K Problems
* (1968) (1976) (1987) (1987) (19921998) (1992) (19922009) (1992) (1994) (1994) (MOSFET 1994) (1996) (MOS and CMOS1998) (1998) (1998) (2000) (2000)
* (2002) (2004) (Hilbert Huang Transform2004) (2004) (2004) (2006) (2006) (2006) (2006) (2006) (2010) (2012) (2012) (2012)31 11%72004