#include <otbWaveletFilterBank.h>

Inheritance diagram for otb::WaveletFilterBank< TInputImage, TOutputImage, TWaveletOperator, TDirectionOfTransformation >:

Collaboration diagram for otb::WaveletFilterBank< TInputImage, TOutputImage, TWaveletOperator, TDirectionOfTransformation >:

Public Types
typedef itk::SmartPointer< const Self >	ConstPointer

typedef itk::SmartPointer< Self >	Pointer

typedef WaveletFilterBank	Self

typedef itk::ImageToImageFilter< TInputImage, TOutputImage >	Superclass

Public Member Functions
virtual ::itk::LightObject::Pointer	CreateAnother (void) const

virtual const char *	GetNameOfClass () const

Static Public Member Functions
static Pointer	New ()

Protected Member Functions
	WaveletFilterBank ()

virtual	~WaveletFilterBank ()

Private Member Functions
void	operator= (const Self &)

	WaveletFilterBank (const Self &)

Detailed Description

template<class TInputImage, class TOutputImage, class TWaveletOperator, Wavelet::WaveletDirection TDirectionOfTransformation>
class otb::WaveletFilterBank< TInputImage, TOutputImage, TWaveletOperator, TDirectionOfTransformation >

One level stationary wavelet transform.

Template specialization of FilterBank for inverse transformation.

Template specialization of FilterBank for forward transformation.

This implementation performs a low-pass / high-pass wavelet transformation of an image. The wavelet transformation is defined by a inner product (ie. convolution-like operation).

the inner operator are supposed to be defined through 1D filters. Then, the forward transformation yields $2^{\text{Dim}}$ output images, while the inverse transformation requires $2^{\text{Dim}}$ input image for one output.

In case of 1D, GetOutput(0) -> LowPass

           GetOutput(1) -> HighPass

In case of 2D, Line (Dim 1) Col (Dim 0)

           GetOutput(0) -> LowPass,      LowPass

           GetOutput(1) -> LowPass,      HighPass

           GetOutput(2) -> HighPass,     LowPass

           GetOutput(3) -> HighPass,     HighPass

In case of nD data, assume x_n=0 stands for LowPass and x_n=1 stands for HighPass at a give dimension n. Then

GetOutput( x_(n-1) << (n-1) + x_(n-2) << (n-2) + ... + x_1 << 1 + x_0 )

Dim (n-1)   Dim (n-2)   Dim (n-3)   ...   Dim 1  Dim 0

-> x_(n-1) x_(n-2) x_(n-3) x_1 x_0

And conversely in the inverse transformation.

Todo:: : At present version, there is not consideration on meta data information that can be transmitted from the input(s) to the output(s)...

The two choice (Wavelet::FORWARD/Wavelet::INVERSE) yield specific implementation of the templates (header redeclaration is given at bottom of otbWaveletFilterBank.h for the Wavelet::INVERSE case)

See also: WaveletOperator

This implementation performs a low-pass / high-pass wavelet transformation of an image. The wavelet transformation is defined by a inner product (ie. convolution-like operation).

The inner operator are supposed to be defined through 1D filters. Then, the forward transformation yields $2^{\text{Dim}}$ output images, while the inverse transformation requires $2^{\text{Dim}}$ input image for one output.

In case of 1D, GetOutput(0) -> LowPass

           GetOutput(1) -> HighPass

In case of 2D, Line (Dim 1) Col (Dim 0)

           GetOutput(0) -> LowPass,      LowPass

           GetOutput(1) -> LowPass,      HighPass

           GetOutput(2) -> HighPass,     LowPass

           GetOutput(3) -> HighPass,     HighPass

In case of nD data, assume x_n=0 stands for LowPass and x_n=1 stands for HighPass at a give dimension n. Then

GetOutput( x_(n-1) << (n-1) + x_(n-2) << (n-2) + ... + x_1 << 1 + x_0 )

Dim (n-1)   Dim (n-2)   Dim (n-3)   ...   Dim 1  Dim 0

-> x_(n-1) x_(n-2) x_(n-3) x_1 x_0

And conversely in the inverse transformation.

Todo:: : At present version, there is not consideration on meta data information that can be transmitted from the input(s) to the output(s)...

See also: WaveletOperator

This implementation performs a low-pass / high-pass wavelet transformation of an image. The wavelet transformation is defined by a inner product (ie. convolution-like operation).

The inner operator are supposed to be defined through 1D filters. Then, the forward transformation yields $2^{\text{Dim}}$ output images, while the inverse transformation requires $2^{\text{Dim}}$ input image for one output.

In case of 1D, GetOutput(0) -> LowPass

           GetOutput(1) -> HighPass

In case of 2D, Line (Dim 1) Col (Dim 0)

           GetOutput(0) -> LowPass,      LowPass

           GetOutput(1) -> LowPass,      HighPass

           GetOutput(2) -> HighPass,     LowPass

           GetOutput(3) -> HighPass,     HighPass

In case of nD data, assume x_n=0 stands for LowPass and x_n=1 stands for HighPass at a give dimension n. Then

GetOutput( x_(n-1) << (n-1) + x_(n-2) << (n-2) + ... + x_1 << 1 + x_0 )

Dim (n-1)   Dim (n-2)   Dim (n-3)   ...   Dim 1  Dim 0

-> x_(n-1) x_(n-2) x_(n-3) x_1 x_0

And conversely in the inverse transformation.

Todo:: : At present version, there is not consideration on meta data information that can be transmitted from the input(s) to the output(s)...

See also: WaveletOperator

Definition at line 87 of file otbWaveletFilterBank.h.