image_processing.h

Go to the documentation of this file.

/**
 *  \file IMP/em2d/image_processing.h
 *  \brief Image processing functions
 *
 *  Copyright 2007-2022 IMP Inventors. All rights reserved.
*/

#ifndef IMPEM2D_IMAGE_PROCESSING_H
#define IMPEM2D_IMAGE_PROCESSING_H

#include <IMP/em2d/em2d_config.h>
#include "IMP/em2d/PolarResamplingParameters.h"
#include "IMP/em2d/opencv_interface.h"
#include "IMP/base_types.h"
#include <algorithm>
#include <cmath>
#include <cereal/access.hpp>

IMPEM2D_BEGIN_NAMESPACE

//! Class to provide all the parameters to the segmentation function
class IMPEM2DEXPORT SegmentationParameters {
private:
  friend class cereal::access;

  template<class Archive> void serialize(Archive &ar) {
    ar(image_pixel_size, diffusion_beta, diffusion_timesteps,
       fill_holes_stddevs, opening_kernel, remove_sizing_percentage,
       binary_background, binary_foreground, threshold);
  }

 public:
  double image_pixel_size;
  double diffusion_beta;
  double diffusion_timesteps;
  double fill_holes_stddevs;
  cv::Mat opening_kernel;
  double remove_sizing_percentage;
  int binary_background;
  int binary_foreground;
  double threshold;  // values of the image below the threshold will be set to 0

  SegmentationParameters()
      : image_pixel_size(1),
        diffusion_beta(45),
        diffusion_timesteps(200),
        fill_holes_stddevs(1.0),
        remove_sizing_percentage(0.1),
        binary_background(0),
        binary_foreground(1),
        threshold(0.0) {
    opening_kernel = cv::Mat::ones(3, 3, CV_64FC1);
  }

  SegmentationParameters(double apix, double diff_beta,
                         unsigned int diff_timesteps, double fh_stddevs,
                         const cv::Mat &kr, int background, int foreground)
      : image_pixel_size(apix),
        diffusion_beta(diff_beta),
        diffusion_timesteps(diff_timesteps),
        fill_holes_stddevs(fh_stddevs),
        opening_kernel(kr),
        binary_background(background),
        binary_foreground(foreground) {};

  void show(std::ostream &out = std::cout) const {
    out << "Diffusion parameters: " << std::endl;
    out << "image_pixel_size " << image_pixel_size << std::endl;
    out << "diffusion_beta " << diffusion_beta << std::endl;
    out << "diffusion_timesteps " << diffusion_timesteps << std::endl;
    out << "fill_holes_stddevs " << fill_holes_stddevs << std::endl;
    out << "Opening kernel " << opening_kernel.rows << "x"
        << opening_kernel.cols << std::endl;
    out << "binary_background " << binary_background << std::endl;
    out << "binary_foreground " << binary_foreground << std::endl;
  }
};
IMP_VALUES(SegmentationParameters, SegmentationParametersList);

/*! Information for the match of a template in one image. The pair is the
    pixel of the image where the match is found.
*/
class IMPEM2DEXPORT MatchTemplateResult {
 public:
  IntPair pair;
  double cross_correlation;
  MatchTemplateResult(IntPair p, double cc) : pair(p), cross_correlation(cc) {}

  MatchTemplateResult() {}

  void show(std::ostream &out = std::cout) const {
    out << "MatchTemplateResult: Pair (" << pair.first << "," << pair.second
        << ") ccc = " << cross_correlation << std::endl;
  }
private:
  friend class cereal::access;

  template<class Archive> void serialize(Archive &ar) {
    ar(pair.first, pair.second, cross_correlation);
  }
};
IMP_VALUES(MatchTemplateResult, MatchTemplateResults);

//! Applies a binary mask to an image.
/*!
  \param[in] m Input matrix
  \param[in] result matrix with the result
  \param[in] mask If the mask is 0, the result has the value of val. Otherwise
             is the value of m.
  \param[in] val value to apply when the mask is 0
*/
IMPEM2DEXPORT void apply_mask(const cv::Mat &m, cv::Mat &result,
                              const cvIntMat &mask, double val);

//! Applies a circular to a matrix. The center of the mask is the center of the
//! matrix.
IMPEM2DEXPORT void apply_circular_mask(const cv::Mat &mat, cv::Mat &result,
                                       int radius, double value = 0.0);

/*! Mean of an matrix inside a mask
  \param[in] mat The matrix
  \param[in] mask If a pixel is not zero, is considered for computing the mean
  \returns The mean
*/
IMPEM2DEXPORT double get_mean(const cv::Mat &mat, const cvIntMat &mask);

/*! Creates a circular matrix that is a mask
  \param[in] rows
  \param[in] cols
  \param[in] radius The radius of the mask. The center of the mask is the
             center of the matrix
  \returns A binary matrix (0/1) with value 1 inside the radius
*/
IMPEM2DEXPORT cvIntMat create_circular_mask(int rows, int cols, int radius);

//! Removes small objects from a matrix of integers.
/*!
  \param[in] m the matrix
  \param[in] percentage The percentage respect to the largest object that
             other objects have to be in order to survive the removal.
  \param[in] background value for the background after removed
  \param[in] foreground value for the foreground after removed
*/
IMPEM2DEXPORT void do_remove_small_objects(cvIntMat &m, double percentage,
                                           int background = 0,
                                           int foreground = 1);

IMPEM2DEXPORT void do_histogram_stretching(cv::Mat &m, int boxes, int offset);

//! (U. Adiga, 2005)
/*!
  \param[in] m binary matrix to dilate and shrink
  \param[in] grayscale grayscale matrix that controls the shrinking
  \param[in] kernel dilation kernel
  \note Only tested with binary matrices m with background =0 and foreground = 1
*/
void IMPEM2DEXPORT do_dilate_and_shrink_warp(cv::Mat &m,
                                             const cv::Mat &grayscale,
                                             cv::Mat &kernel);

//! morphologic grayscale reconstruction (L Vincent, 1993)
/*!
  \param[in] mask image to reconstruct
  \param[out] marker this image contains the initial marker points and will
              contain the final result
  \param[in] neighbors_mode number of neighbors for a pixel to consider when
             doing the morphologic reconstruction (values: 4, 8).
*/
void IMPEM2DEXPORT do_morphologic_reconstruction(const cv::Mat &mask,
                                                 cv::Mat &marker,
                                                 int neighbors_mode = 4);

//! Labeling function for a matrix
/*!
  \param[in] m binary matrix to scan. The matrix needs to contain zeros and
             ones but they can be stored as doubles, floats or ints
  \param[out] mat_to_label matrix it is returned as a matrix of ints
  \return labels The number of labels in the image
*/
IMPEM2DEXPORT int do_labeling(const cvIntMat &m, cvIntMat &mat_to_label);

//! Segmentation of images
/*!
  \param[in] m The EM image to segment
  \param[in] result The segmented image, with the shape of the molecule
  \param[in] params
*/
IMPEM2DEXPORT void do_segmentation(const cv::Mat &m, cv::Mat &result,
                                   const SegmentationParameters &params);

//! Smoothing filter by application of the reaction-diffusion
//! equation of Beltrami flow. Adiga, JSB, 2005
/*!
  `beta` is the contribution of diffusion versus edge enhancement:
              0 - pure reaction, 90 - pure diffusion
  \note The function only works for matrices containing doubles
*/
IMPEM2DEXPORT void apply_diffusion_filter(const cv::Mat &m, cv::Mat &result,
                                          double beta, double pixelsize,
                                          unsigned int time_steps);

//! Partial derivative with respect to time for an image filtered with
//! diffusion-reaction
/*!
  \param[in] m input matrix
  \param[in] der output derivative matrix
  \param[in] dx - step for x
  \param[in] dy - step for y
  \param[in] ang - parameter for weight diffusion and edge detection (90-0)
*/
IMPEM2DEXPORT void get_diffusion_filtering_partial_derivative(
    const cv::Mat &m, cv::Mat &der, double dx, double dy, double ang);

//! Fills the holes in the matrix m of height h
/*!
  \param[in] m the input matrix
  \param[in] result the result matrix
  \param[in] h the height
  \note The function does not work in-place
*/
IMPEM2DEXPORT void do_fill_holes(const cv::Mat &m, cv::Mat &result, double h);

//! Gets the domes of m with height h
/*!
  \param[in] m the input matrix
  \param[in] result the result matrix
  \param[in] h the height
  \note The function does not work in-place
*/
IMPEM2DEXPORT void get_domes(cv::Mat &m, cv::Mat &result, double h);

//! Combines the fill holes and thresholding operations together with normalize
/*!
  \param[in] m Input matrix
  \param[out] result the result matrix
  \param[in] n_stddevs Number of standard deviations used to compute the holes
  \param[in] threshold Removes values below the threshold value
  \note The function does normalize -> fill_holes -> normalize -> threshold ->
    normalize
*/
IMPEM2DEXPORT void do_combined_fill_holes_and_threshold(cv::Mat &m,
                                                        cv::Mat &result,
                                                        double n_stddevs,
                                                        double threshold = 0.0);

//! Computes the histogram of a matrix.
/*!
  \param[in] m Matrix with the data
  \param[in] bins Number of bins to use in the histogram
  \return vector with the values for each bin
*/
IMPEM2DEXPORT Floats get_histogram(const cv::Mat &m, int bins);

//! Variance filter for an image. Computes the variance for each pixel using
//! the surrounding ones.
/*!
 \param[in] input image with the data
 \param[out] filtered matrix result of the filtering with the variances
 \param[in] kernelsize The variance is computed using kernelsize x kernelsize
             pixels around each one. Kernelsize can only be odd.
*/
IMPEM2DEXPORT void apply_variance_filter(const cv::Mat &input,
                                         cv::Mat &filtered, int kernelsize);

//!Add noise to the values of a matrix.
/*!
    Supported distributions:
   - uniform distribution, giving the range (lower, upper). DEFAULT
   - Gaussian distribution, giving the mean and the standard deviation
   \code
   add_noise(v1,0, 1);
   // uniform distribution between 0 and 1

   v1.add_noise(0, 1, "uniform");
   // the same

   v1.add_noise(0, 1, "gaussian");
  // Gaussian distribution with 0 mean and stddev=1

   \endcode
*/
IMPEM2DEXPORT void add_noise(cv::Mat &v, double op1, double op2,
                             const String &mode = "uniform", double df = 3);

//! Resamples a matrix to polar coordinates.
/*!
  \param[in] input matrix to resample
  \param[out] resampled result matrix to contain the resampling
  \param[in] polar_params Parameters used for the resampling. Extremely useful
            for speeding up the procedure if they are given with the
            transformation maps, that can be built in the
            PolarResamplingParameters class
*/
IMPEM2DEXPORT void do_resample_polar(
    const cv::Mat &input, cv::Mat &resampled,
    const PolarResamplingParameters &polar_params);

//! Normalize a openCV matrix to mean 0 and stddev 1. It is done in place
IMPEM2DEXPORT void do_normalize(cv::Mat &m);

//! Applies a transformation to a matrix. First rotates the matrix using the
//! matrix center as the origin of the rotation, and then applies
//! the translation
IMPEM2DEXPORT void get_transformed(const cv::Mat &input, cv::Mat &transformed,
                                   const algebra::Transformation2D &T);

/*! Extends the borders of an image
 \param[in] orig The image to extend
 \param[in] dst The image destination
 \param[in] pix number of pixels to extend the borders
*/
IMPEM2DEXPORT void do_extend_borders(cv::Mat &orig, cv::Mat &dst,
                                     unsigned int pix);

/*! Applies a threshold to an image
  `threshold` is a value such that all pixels below this value are set to zero
*/
IMPEM2DEXPORT void apply_threshold(cv::Mat &m, cv::Mat &result,
                                   double threshold = 0.0);

/*! morphologic enhancement of the contrast
  This function detects areas in the images and enhances the contrast between
  them
  \param[in] m Input matrix
  \param[out] result
  \param[in] kernel morphologic kernel to use
  \param[in] iterations Higher number, more contrast
*/
IMPEM2DEXPORT void do_morphologic_contrast_enhancement(const cv::Mat &m,
                                                       cv::Mat &result,
                                                       const cv::Mat &kernel,
                                                       unsigned int iterations);
/*! Morphologic gradient: dilation-erosion
  \param[in] m Input matrix
  \param[out] result
  \param[in] kernel morphologic kernel
*/
IMPEM2DEXPORT void get_morphologic_gradient(const cv::Mat &m, cv::Mat &result,
                                            const cv::Mat &kernel);

/*! Get the percentage of overlap between two matrices.
  Two images are overlapping in a pixel if both have values > 0.
  The overlap is (pixels overlapping) / (pixels > 0 in m2)

  \param[in] m1 First matrix
  \param[in] m2 Matrix used to check the overlap. This matrix can be of the
            same size of m1 or smaller.
  \param[in] center Indicates the position of m1 where to put
             the center m2. E.g., if center is (32,16) the center
             of m2 will be in the pixel (32,16) of m1.
*/
IMPEM2DEXPORT double get_overlap_percentage(cv::Mat &m1, cv::Mat &m2,
                                            const IntPair &center);

/*! Gets the n first matches between an image and a template
  \param[in] m Matrix
  \param[in] templ Matrix with a template to be found in m
  \param[in] n Number of positions to recover
*/
IMPEM2DEXPORT MatchTemplateResults
    get_best_template_matches(const cv::Mat &m, const cv::Mat &templ,
                              unsigned int n);

/*! Crop an image
  \param[in] m Matrix to crop
  \param[in] center The pixel used as the center for cropping
  \param[in] size The size of the new image
  \return A matrix with the cropped region
*/
IMPEM2DEXPORT cv::Mat crop(const cv::Mat &m, const IntPair &center, int size);

IMPEM2D_END_NAMESPACE

#endif /* IMPEM2D_IMAGE_PROCESSING_H */