! Await great changes to happen

Preprocessing

Fingerprint preprocessing module for DBOX

Dependencies:

• OpenCV 4
• Caffe 1
• ArrayFire
• CUDA 8-10.2
• cuDNN up to 7.6.5
• Qt 5 + Qt Creator

Getting Started:

1. You need to provide valid paths to these libraries and their header files in .pro file.
2. Build and run the project to generate .so (.dll / .lib) files
3. Include the library and header files to your own application
4. Copy the 'core' folder to your root project directory

API

Required

int loadInput(cv::Mat imgOriginal);
int loadInput(QVector<cv::Mat> imgOriginals);
int loadInput(QString inputPath);
  
void start();  

Usage:

1. First you have to load input, which can be:

• cv::Mat image
• QVector of cv::Mat images
• Image input path for a .jpg, .png, .bmp, .tif file
• Image input directory

2. Just call start()

Optional

void setPreprocessingParams(int blockSize = 13, double gaborLambda = 9, double gaborSigma = 3, int gaussBlockBasic = 1, double gaussSigmaBasic = 1.0, int gaussBlockAdvanced = 121, double gaussSigmaAdvanced = 10.0, int holeSize = 20);  
  
void setFeatures(bool useAdvancedMode, bool useContrastEnhancement = true, bool useAdvancedOrientationMap = true, bool useHoleRemover = true, bool generateInvertedSceleton = true, bool useQualityMap = true, bool useMask = false, bool useFrequencyMap = false);  

void setCPUOnly(bool enabled, int numThreads = 0);  
  
void setMaskParams(CAFFE_FILES maskFiles, int blockSize, int exBlockSize, bool useSmooth);  
  
void setFrequencyMapParams(CAFFE_FILES freqFiles, int blockSize, int exBlockSize);  

Usage:

• With setPreprocessingParams(...) you can set the must important parameters required for image processing, the default parameters are recommended for 500dpi images
• With setFeatures(...) you can set some optional features which can affect the output image quality, the advanceMode gives you all the intermediate results generated during preprocessing
• With setCPUOnly(...) you can force the library to use only CPU for image processing (notice: it can be slower than GPU) and set the number of CPU threads you want to use for image processing in defined phases (notice: numThreads = 0 means automatic ideal thread number)
• With setMaskParams(...) and setFrequencyMapParams(...) you can set the Caffe model files and parameteres required for classification with neural network to generate mask or frequency map for the input image

SIGNALS:

preprocessingDoneSignal(PREPROCESSING_ALL_RESULTS results);  
  
preprocessingDoneSignal(PREPROCESSING_RESULTS results);  
  
preprocessingSequenceDoneSignal(QMap<QString, PREPROCESSING_ALL_RESULTS> results);  
  
preprocessingSequenceDoneSignal(QMap<QString, PREPROCESSING_RESULTS> results);
  
preprocessingDurationSignal(PREPROCESSING_DURATIONS durations);  
  
preprocessingErrorSignal(int errorcode);  

Important notice:

• If you load an image or an image path
  • You get preprocessingDoneSignal with PREPROCESSING_ALL_RESULTS if the advancedMode is enabled
  • You get preprocessingDoneSignal with PREPROCESSING_RESULTS if the advancedMode is disabled
• If you load a vector with images or an input directory
  • You get preprocessingSequenceDoneSignal with PREPROCESSING_ALL_RESULTS if the advancedMode is enabled
  • You get preprocessingSequenceDoneSignal with PREPROCESSING_RESULTS if advancedMode is disabled
• You get preprocessingDurationSignal with duration values in ms for each phase during preprocessing if it's finished successfully
• You get preprocessingErrorSignal with the error code if an error occured during preprocessing

A simple example how to use signals in your application
yourclass.h:

#include "preprocessing.h"

class YourClass: public QObject
{
    Q_OBJECT  
  
private:  
    Preprocessing p;  
    
private slots:
    void proprocessingResultsSlot(PREPROCESSING_RESULTS result);
}

yourclass.cpp:

#include "yourclass.h"
YourClass::YourClass()
{
    qRegisterMetaType<PREPROCESSING_RESULTS >("PREPROCESSING_RESULTS");
    connect(&p, SIGNAL(preprocessingDoneSignal(PREPROCESSING_RESULTS)), this, SLOT(proprocessingResultsSlot(PREPROCESSING_RESULTS)));
}

void YourClass::proprocessingResultsSlot(PREPROCESSING_RESULTS result)
{
    cv::imshow("Fingerprint Skeleton", result.imgSkeleton);
}

For more please visit Qt Signals & Slots.