ImFusion SDK 4.3
SimilarityMeasureMI_CPU Class Reference

#include <ImFusion/Reg/SimilarityMeasureMI.h>

CPU implementation of Mutual Information. More...

+ Inheritance diagram for SimilarityMeasureMI_CPU:

Detailed Description

CPU implementation of Mutual Information.

Besides a generic implementation for any number of bins and input images, an optimized version is provided for specific image types and number of bins. Currently optimized are:

  • unsigned char, 256 bins, 256 joint bins
  • unsigned short, 256 bins, 256 joint bins
  • unsigned short, 1024 bins, 1024 joint bins
  • unsigned short, 4096 bins, 1024 joint bins
  • unsigned short, 4096 bins, 4096 joint bins
  • unsigned short, 65536 bins, 256 joint bins
  • unsigned short, 65536 bins, 4096 joint bins

Public Member Functions

 SimilarityMeasureMI_CPU (SharedImage *img1, SharedImage *img2)
 
double compute (const ImageTransformer *t, size_t *overlap=nullptr) override
 Computes and returns the similarity measure.
 
- Public Member Functions inherited from SimilarityMeasureMI
 SimilarityMeasureMI (SharedImage *img1, SharedImage *img2)
 
void setImage1Bins (int bins)
 Sets the number of bins the first images histogram.
 
int image1Bins () const
 
void setImage2Bins (int bins)
 Sets the number of bins the second images histogram.
 
int image2Bins () const
 
SharedImagehistogramJoint ()
 Returns the joint histogram of both images.
 
void configure (const Properties *p) override
 Set one or multiple properties.
 
void configuration (Properties *p) const override
 Retrieve the properties of this object.
 
double computeEntropy (const TypedImage< unsigned int > &histogram) const
 Computes the Shanon entropy using the histogram.
 
bool supportsWeights () const override
 Queries the capability whether this similarity measure supports pixel-wise weights.
 
- Public Member Functions inherited from SimilarityMeasureImpl
 SimilarityMeasureImpl (SharedImage *img1, SharedImage *img2)
 Sets the two images that should be compared. Does not transfer ownership.
 
virtual bool setImage1 (SharedImage *img1)
 Replaces the image1 that was set in the constructor with the given image.
 
virtual bool setImage2 (SharedImage *img2)
 Replaces the image2 that was set in the constructor with the given image.
 
virtual SharedImageimage1 () const
 
virtual SharedImageimage2 () const
 
void enableImage1Mask (bool enable)
 Specify whether a mask on the first image shall be used if available (true by default)
 
void enableImage2Mask (bool enable)
 Specify whether a mask on the second image shall be used if available (true by default)
 
virtual void setWeights (const SharedImage *weights1, const SharedImage *weights2)
 Sets the images to use as weights for the similarity metric.
 
bool image1MaskEnabled () const
 Return the flag for using the mask of the first image.
 
bool image2MaskEnabled () const
 Return the flag for using the mask of the second image.
 
- Public Member Functions inherited from Configurable
virtual void configureDefaults ()
 Retrieve the properties of this object, replaces values with their defaults and sets it again.
 
void registerParameter (ParameterBase *param)
 Register the given Parameter or SubProperty, so that it will be configured during configure()/configuration().
 
void unregisterParameter (const ParameterBase *param)
 Remove the given Parameter or SubProperty from the list of registered parameters.
 
 Configurable (const Configurable &rhs)
 
 Configurable (Configurable &&rhs) noexcept
 
Configurableoperator= (const Configurable &)
 
Configurableoperator= (Configurable &&) noexcept
 

Protected Member Functions

void updateHistogramSize () override
 
- Protected Member Functions inherited from SimilarityMeasureMI
void createDebugTexture ()
 
void updateRanges () const
 Extracts the value range of the images and sets the m_img1_range and m_img2_range values.
 
- Protected Member Functions inherited from SimilarityMeasureImpl
template<typename T>
double callTyped (T *obj, const MemImage &img1, const MemImage &img2, const ImageTransformer *t, size_t *overlap=nullptr)
 Helper method to transform the MemImages to TypedImages.
 
void setMaskIncludeArguments (GL::Program *program)
 Sets all uniforms for SimilarityMeasureMask.glh.
 
void initMaskInclude (GL::Program *program)
 Initializes the code for the SimilarityMeasureMask.glh code.
 
std::string glDefines (bool patchBased) const
 Return shared GL program defines for patch-based and weights, used in LNCC and LC2 OpenGL subclasses.
 

Additional Inherited Members

- Static Public Member Functions inherited from SimilarityMeasureMI
static void defaultConfiguration (Properties *p)
 
- Static Public Member Functions inherited from SimilarityMeasureImpl
static void defaultConfiguration (Properties *p)
 Retrieve the default properties of this class.
 
- Public Attributes inherited from Configurable
Signal signalParametersChanged
 Emitted whenever one of the registered Parameters' or SubPropertys' signalValueChanged signal was emitted.
 
- Protected Attributes inherited from SimilarityMeasureMI
int m_bins1
 
int m_bins2
 
SharedImagem_histogramJoint
 
bool m_exportJointHistogram
 
std::optional< std::pair< double, double > > m_img1Range
 Value range of the first image.
 
std::optional< std::pair< double, double > > m_img2Range
 Value range of the second image.
 
- Protected Attributes inherited from SimilarityMeasureImpl
SharedImagem_img1
 
SharedImagem_img2
 
bool m_useImg1Mask
 
bool m_useImg2Mask
 
std::shared_ptr< GlMask::MultiIncludeTypem_maskImg1
 
std::shared_ptr< GlMask::MultiIncludeTypem_maskImg2
 
const SharedImagem_weights1
 
const SharedImagem_weights2
 
- Protected Attributes inherited from Configurable
std::vector< Paramm_params
 List of all registered Parameter and SubProperty instances.
 

Member Function Documentation

◆ compute()

double compute ( const ImageTransformer * t,
size_t * overlap = nullptr )
overridevirtual

Computes and returns the similarity measure.

The higher the value the higher the similarity between the two images. The returned value does not need to be normalized, therefore values from different similarity measures should not be compared directly. The computation always iterates each pixel of img1 and compares it to the related pixel of img2. If no ImageTransformer is supplied, the image coordinates of a pixel in img1 are used directly to lookup the value in img2, no transformation is assumed. In this case it's usually advisable that img1 and img2 have the same dimensions, but this is not enforced. If an ImageTransformer is given, it will be used to transfer from img1 pixel coordinates to img2 pixel coordinates including any world transformations. In this case the dimensions of the two images can be unrelated. The optional argument overlap contains the overlap of the two images in pixels.

Implements SimilarityMeasureImpl.

◆ updateHistogramSize()

void updateHistogramSize ( )
overrideprotectedvirtual

Implements SimilarityMeasureMI.


The documentation for this class was generated from the following file:
Search Tab / S to search, Esc to close