Scribd
Upload
502 views

Wavelets and Multiresolution Processing PDF

The document discusses wavelet transforms and multiresolution processing. It provides an overview of wavelet transforms as an alternative to Fourier transforms that can provide both spectral and temporal information. It describes how wavelet transforms form the foundation of multiresolution theory, including image pyramids, subband coding, and filter banks. It also discusses multiresolution analysis using scaling functions and wavelets to create approximations of images at different resolutions.

Uploaded by

Ahmad Raza
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
502 views

Wavelets and Multiresolution Processing PDF

The document discusses wavelet transforms and multiresolution processing. It provides an overview of wavelet transforms as an alternative to Fourier transforms that can provide both spectral and temporal information. It describes how wavelet transforms form the foundation of multiresolution theory, including image pyramids, subband coding, and filter banks. It also discusses multiresolution analysis using scaling functions and wavelets to create approximations of images at different resolutions.

Uploaded by

Ahmad Raza
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 15

Wavelets and Multiresolution

Processing
Special Seminar
Shahab Ansari
Wavelet Transform

 An alternative to Fourier transform to compress, transmit and analyze


images
 The basis functions of wavelet transform are wavelets of varying frequency
and finite duration
 These wavelets not only give spectral information but also temporal
information simultaneously
 Wavelet is the foundation of multiresolution theory (Mallat, 1987)
 Subband coding
 Quadrature mirror filtering
 Pyramidal image processing
Wavelet Transform

 Image Pyramids
 Base level is 𝑁 × 𝑁 𝑜𝑟 2 𝐽 × 2 𝐽 where 𝐽 = log 2 𝑁
Wavelet Transform

 Image Pyramids
Wavelet Transform

 Subband Coding
 Filter coefficients
 Filter taps
 Filter order
 FIR filter
Wavelet Transform

 More Filters
Wavelet Transform

 Subband coding
 Two-Band Subband Coding using two filter banks such that 𝑓መ 𝑛 = 𝑓(𝑛)
 For perfect reconstruction (cross-modulated and biorthogonal) filters,
𝑔0 (𝑛) = −1 𝑛 ℎ1 (𝑛)
𝑛+1
𝑔1 (𝑛) = −1 ℎ0 (𝑛)
Wavelet Transform

 Four-Band Subband Coding


Wavelet Transform

 Haar Transform 𝑻 is defined as,


𝑻 = 𝑯𝑭𝑯𝑻
 where 𝑭 is an 𝑁 × 𝑁 image matrix and 𝑯 is an 𝑁 × 𝑁 Haar transformation
matrix containing Haar basis functions
 Haar basis functions is given as,
2𝑃/2 (𝑞 − 1)/2𝑃 ≤ 𝑧 ≤ (𝑞 − 0.5)/2𝑃
1 𝑃
ℎ𝑘 𝑧 = ℎ𝑝𝑞 𝑧 = (𝑞 − 0.5)/2𝑃 ≤ 𝑧 ≤ 𝑞/2𝑃
𝑁 −2
2
0 otherwise, 𝑧 ∈ [0,1]
𝑘 = 2𝑃 + 𝑞 − 1
Wavelet Transform

 Haar Transform
Wavelet Transform

 Multiresolution analysis (MRA)


 A scaling function is used to create a series of approximations of an image each
differing by a factor of 2 in resolution
 The wavelets are used to encode the difference between adjacent
approximations
 Expansion functions (basis functions)
𝑓 𝑥 = ෍ 𝛼𝑘 𝜑𝑘
𝑘

𝜑𝑘 are called the expansion functions and 𝛼𝑘 are the expansion coefficients
Wavelet Transform

 Scaling functions
 Integer translation and binary scaling
𝜑𝑗,𝑘 𝑥 = 2𝑗/2 𝜑(2𝑗 𝑥 − 𝑘)
𝜑𝑗,𝑘 𝑥 ∈ 𝐿2 (𝐑)
Wavelet Transform

 MRA Requirement 1
 The scaling function is orthogonal to its integer translates
 MRA Requirement 2
 The subspaces spanned by the scaling function at low scales are nested within those
spanned at higher scales
 MRA Requirement 3
 The only function that is common to all 𝑉𝑗 is 𝑓 𝑥 = 0
 MRA Requirement 4
 Any function can be represented with arbitrary precision
𝜑 𝑥 = ෍ ℎ𝜑 (𝑛) 2𝜑(2𝑥 − 𝑛)
𝑛
ℎ𝜑 (𝑛) are called scaling function coefficients
Wavelet Transform

 Wavelet functions
𝑗
𝜓𝑗,𝑘 𝑥 = 22 𝜓2𝑗 𝑥 − 𝑘
𝑊𝑗 = Span{𝜓𝑗,𝑘 𝑥 }
𝑘
𝑉𝑗+1 = 𝑉𝑗 ⨁𝑊𝑗
𝜓 𝑥 = ෍ ℎ𝜓 (𝑛) 2𝜑(2𝑥 − 𝑛)
𝑛
Wavelet Transform

 Wavelet functions

You might also like