ImageDev

Measurement Browsing

This example shows how to automatically parse the content of a label analysis and export it in a csv file.
First, a grayscale image is opened and segmented to generate a label image containing 8 objects.

Then, an analysis is performed with three measurements selected. It computes for each object its area, a shape factor, and a set of oriented diameters. The FeretDiameter measurement generates by default an array of 10 diameters corresponding to different orientations with a pitch of 18 degrees.

A first loop shows how to introspect the analysis to get the name of each selected measurement and deploy the array of diameters. A second loop shows how to introspect the analysis to print the measurement results for each object and deploys the diameter distribution. This step demonstrates the ability to browse the content of an analysis without making assumptions about the measurements that have been selected within.

Label Area2d(mm^2) InverseCircularity2d() FeretDiameter2d[0](mm) FeretDiameter2d[1](mm)
1 1058.00 1.23 35.00 38.73
2 1361.00 3.70 47.00 49.31
3 1086.00 1.26 38.00 35.54
4 293.00 1.81 20.00 18.40

In practice, the analysis can be directly converted to a spreadsheet structure with the toDataFrame method. It generates an IOLink DataFrameView object. In Python, this object can be directly printed in the standard output.

Index [label] Area2d InverseCircularity2d ... FeretDiameter2d[direction=9]
0 1058 1.2297213077545166 ... 38.72925567626953
1 1361 3.698363780975342 ... 49.00188064575195
2 1086 1.2601970434188843 ... 42.93851089477539
3 293 1.8089686632156372 ... 22.111291885375977

Finally, IOFormat allows the export of a DataFrameView object in a csv file that can be visualized as an Excel table. This method is more straightforward for exporting an analysis. The previous one gives you more freedom to customize the export for your needs. 

#include <ImageDev/ImageDev.h>
#include <ioformat/IOFormat.h>
#include <string.h>

using namespace imagedev;
using namespace ioformat;
using namespace iolink;

int
main( int argc, char* argv[] )
{
    int status = 0;

    try
    {
        // ImageDev library initialization if not done
        if ( isInitialized() == false )
            imagedev::init();

        // Open a grayscale image from a tif file
        auto imageInput = readImage( std::string( IMAGEDEVDATA_IMAGES_FOLDER ) + "objects.tif" );

        // Threshold and label the binary input
        auto imageBin = thresholdingByCriterion( imageInput,
                                                 ThresholdingByCriterion::ComparisonCriterion::GREATER_THAN_OR_EQUAL_TO,
                                                 40 );
        auto imageLab = labeling2d( imageBin, Labeling2d::LABEL_8_BIT, Labeling2d::CONNECTIVITY_8 );

        // Calibrate this image to match 1 pixel to 1.4 mm
        Vector3d spacing{ 1.4, 1.4, 1 };
        imageLab->setSpatialSpacing( spacing );
        imageLab->setSpatialUnit( "mm" );

        // Define the analysis features to be computed
        AnalysisMsr::Ptr analysis = std::make_shared< AnalysisMsr >();
        analysis->select( NativeMeasurements::area2d );
        analysis->select( NativeMeasurements::inverseCircularity2d );
        analysis->select( NativeMeasurements::feretDiameter2d );

        // Launch the feature extraction on the segmented image
        labelAnalysis( imageLab, imageInput, analysis );

        // Print the analysis table header
        std::string lineToPrint( "Label\t" );
        for ( const auto& measure : analysis->getMeasurements() )
        {
            // Build and print the table header
            if ( measure->shape().size() == 1 )
                // The measurement is a scalar value
                lineToPrint += measure->name() + "(" + measure->information().physicalUnit() + ")\t";
            else if ( measure->shape().size() == 2 )
                // The measurement is an array, loop on it
                for ( size_t j = 0; j < measure->shape()[1]; ++j )
                    lineToPrint += measure->name() + "[" + std::to_string( j ) + "](" +
                                   measure->information().physicalUnit() + ")\t";
        }
        std::cout << lineToPrint << std::endl;

        VectorXu64 index;
        // Print all measurement results for each label
        for ( int i = 0; i < analysis->labelCount(); ++i )
        {
            lineToPrint = std::to_string( i + 1 ) + "\t";
            for ( const auto& measure : analysis->getMeasurements() )
            {
                index = measure->shape();
                index[0] = i;
                if ( measure->shape().size() == 1 )
                    // The measurement is a scalar value
                    lineToPrint += std::to_string( measure->toDouble( index ) ) + "\t";
                else if ( measure->shape().size() == 2 )
                    // The measurement is an array, loop on it
                    for ( size_t j = 0; j < measure->shape()[1]; ++j )
                    {
                        index[1] = j;
                        lineToPrint += std::to_string( measure->toDouble( index ) ) + "\t";
                    }
            }
            std::cout << lineToPrint << std::endl;
        }

        // Export the analysis in a dataframe and save it in a csv file
        auto dataframe = analysis->toDataFrame();
        writeView( dataframe, "T04_03_analysis.csv" );

        std::cout << "This example ran successfully." << std::endl;
    }
    catch ( const imagedev::Exception& error )
    {
        // Print potential exception in the standard output
        std::cerr << "T04_03_MeasurementBrowsing exception: " << error.what() << std::endl;
        status = -1;
    }

    // ImageDev library finalization
    imagedev::finish();

    // Check if we must ask for an enter key to close the program
    if ( !( ( argc == 2 ) && strcmp( argv[1], "--no-stop-at-end" ) == 0 ) )
        std::cout << "Press Enter key to close this window." << std::endl, getchar();

    return status;
}


See also