Loading [MathJax]/jax/output/CommonHTML/jax.js
ImageDev

CudaCrossCorrelation2d

Performs a correlation by convolution between a two-dimensional grayscale image and a grayscale kernel. The calculations are performed on the GPU.

Access to parameter description

This command is experimental, his signature may be modified between now and his final version.

 For an introduction: section Image Correlation.

This algorithm performs a cross-correlation, or correlation with a mutiplication metric, between a gray level image I and a gray level kernel K, returning the correlation image O.

The different possibilities for selecting a correlation peak are presented below using a 1-D correlation between an image and kernel. In the image, the kernel appears 6 times with different contrast and luminosity.
<b> Figure 1.</b> 1D image and kernel
Figure 1. 1D image and kernel

The 6 examples show the kernel appearing with different contrast and luminosity.
<b> Figure 2.</b> Examples of possible 1D correlation
Figure 2. Examples of possible 1D correlation


The cross-correlation metric is computed in accordance with the CorrelationMode parameter.

DIRECT Correlation Mode

O(n,m)=kx∑i=1ky∑j=1K(i,j)×I(n+i−kx2,m+j−ky2) In this mode, 3 of the 6 patterns matching the kernel are detected. These are only high luminosity patterns. The best matching is obtained for the highest contrast and luminosity pattern.

<b> Figure 3.</b> Example of 1D direct cross-correlation
Figure 3. Example of 1D direct cross-correlation

MEAN Correlation Mode

O(n,m)=kx∑i=1ky∑j=1(K(i,j)−μ(K))×(I(n+i−kx2,m+j−ky2)−μ(I)(n,m)) In this mode, 2 of the 6 patterns matching the kernel are detected. These are only high contrast patterns. The confidence rate is the same for the two 2 patterns.

<b> Figure 4.</b> Example of 1D mean cross-correlation
Figure 4. Example of 1D mean cross-correlation

Note: This algorithm returns the main correlation peak in the outputMeasurement object. More correlation peaks can be extracted from the outputImage correlation image with the LocalMaxima2d algorithm

See also

Function Syntax

This function returns a CudaCrossCorrelation2dOutput structure containing outputImage and outputMeasurement. 
// Output structure of the cudaCrossCorrelation2d function.
struct CudaCrossCorrelation2dOutput
{
    /// The output correlation image. Its dimensions are forced to the same values as the input. Its data type is forced to floating point.
    std::shared_ptr< iolink::ImageView > outputImage;
    /// The correlation matching results.
    CrossCorrelation2dMsr::Ptr outputMeasurement;
};

// Function prototype
CudaCrossCorrelation2dOutput
cudaCrossCorrelation2d( std::shared_ptr< iolink::ImageView > inputImage,
                        std::shared_ptr< iolink::ImageView > inputKernelImage,
                        CudaCrossCorrelation2d::CorrelationMode correlationMode,
                        CudaCrossCorrelation2d::TilingMode tilingMode,
                        iolink::Vector2u32 tileSize,
                        CudaContext::Ptr cudaContext,
                        std::shared_ptr< iolink::ImageView > outputImage = nullptr,
                        CrossCorrelation2dMsr::Ptr outputMeasurement = nullptr );

Class Syntax

Parameters

Parameter Name Description Type Supported Values Default Value
input
inputImage
The input grayscale image. Image Grayscale nullptr
input
inputKernelImage
The correlation kernel. Image Grayscale nullptr
input
correlationMode
The normalization mode for correlation.
DIRECT Direct correlation (no normalization).
MEAN Mean normalized correlation (luminosity).
Enumeration DIRECT
input
tilingMode
The way to manage the GPU memory.
NONE The entire input image is transferred to the GPU memory. If the total input, intermediate and output data size exceed the GPU memory, the computation will fail.
USER_DEFINED The input image is processed by tiles of a predefined size.
Enumeration NONE
input
tileSize
The tile width and height in pixels. They must be greater than or equal to the correlation kernel width and height. This parameter is used only in USER_DEFINED tiling mode. Vector2u32 Any value {1024, 1024}
input
cudaContext
CUDA context information. CudaContext nullptr
output
outputImage
The output correlation image. Its dimensions are forced to the same values as the input. Its data type is forced to floating point. Image nullptr
output
outputMeasurement
The correlation matching results. CrossCorrelation2dMsr nullptr

Object Examples

auto polystyrene = ioformat::readImage( std::string( IMAGEDEVDATA_IMAGES_FOLDER ) + "polystyrene.tif" );

CudaCrossCorrelation2d cudaCrossCorrelation2dAlgo;
cudaCrossCorrelation2dAlgo.setInputImage( polystyrene );
cudaCrossCorrelation2dAlgo.setInputKernelImage( polystyrene );
cudaCrossCorrelation2dAlgo.setCorrelationMode( CudaCrossCorrelation2d::CorrelationMode::DIRECT );
cudaCrossCorrelation2dAlgo.setTilingMode( CudaCrossCorrelation2d::TilingMode::NONE );
cudaCrossCorrelation2dAlgo.setTileSize( {264, 264} );
cudaCrossCorrelation2dAlgo.setCudaContext( nullptr );
cudaCrossCorrelation2dAlgo.setOutputImage( iolink::ImageViewFactory::allocate( iolink::VectorXu64( { 1, 1 } ), iolink::DataTypeId::UINT8 ) );
cudaCrossCorrelation2dAlgo.execute();

std::cout << "outputImage:" << cudaCrossCorrelation2dAlgo.outputImage()->toString();
std::cout << "matchingPositionX: " << cudaCrossCorrelation2dAlgo.outputMeasurement()->matchingPositionX( 0 ) ;

Function Examples

auto polystyrene = ioformat::readImage( std::string( IMAGEDEVDATA_IMAGES_FOLDER ) + "polystyrene.tif" );

auto result = cudaCrossCorrelation2d( polystyrene, polystyrene, CudaCrossCorrelation2d::CorrelationMode::DIRECT, CudaCrossCorrelation2d::TilingMode::NONE, {264, 264}, nullptr , iolink::ImageViewFactory::allocate( iolink::VectorXu64( { 1, 1 } ), iolink::DataTypeId::UINT8 ));

std::cout << "outputImage:" << result.outputImage->toString();
std::cout << "matchingPositionX: " << result.outputMeasurement->matchingPositionX( 0 ) ;