Method Parameters

src/depth_packet_processor.cpp:37–98

{

DepthPacketProcessor::Parameters::Parameters()
{
  ab_multiplier = 0.6666667f;
  ab_multiplier_per_frq[0] = 1.322581f;
  ab_multiplier_per_frq[1] = 1.0f;
  ab_multiplier_per_frq[2] = 1.612903f;
  ab_output_multiplier = 16.0f;

  phase_in_rad[0] = 0.0f;
  phase_in_rad[1] = 2.094395f;
  phase_in_rad[2] = 4.18879f;

  joint_bilateral_ab_threshold = 3.0f;
  joint_bilateral_max_edge = 2.5f;
  joint_bilateral_exp = 5.0f;

  gaussian_kernel[0] = 0.1069973f;
  gaussian_kernel[1] = 0.1131098f;
  gaussian_kernel[2] = 0.1069973f;
  gaussian_kernel[3] = 0.1131098f;
  gaussian_kernel[4] = 0.1195716f;
  gaussian_kernel[5] = 0.1131098f;
  gaussian_kernel[6] = 0.1069973f;
  gaussian_kernel[7] = 0.1131098f;
  gaussian_kernel[8] = 0.1069973f;

  phase_offset = 0.0f;
  unambigious_dist = 2083.333f;
  individual_ab_threshold  = 3.0f;
  ab_threshold = 10.0f;
  ab_confidence_slope = -0.5330578f;
  ab_confidence_offset = 0.7694894f;
  min_dealias_confidence = 0.3490659f;
  max_dealias_confidence = 0.6108653f;

  edge_ab_avg_min_value = 50.0f;
  edge_ab_std_dev_threshold = 0.05f;
  edge_close_delta_threshold = 50.0f;
  edge_far_delta_threshold = 30.0f;
  edge_max_delta_threshold = 100.0f;
  edge_avg_delta_threshold = 0.0f;
  max_edge_count  = 5.0f;

/*
 * These are parameters for the method described in "Efficient Phase Unwrapping
 * using Kernel Density Estimation", ECCV 2016, Felix Järemo Lawin, Per-Erik Forssen and
 * Hannes Ovren, see http://www.cvl.isy.liu.se/research/datasets/kinect2-dataset/.
 */

  kde_sigma_sqr = 0.0239282226563f; //the scale of the kernel in the KDE, h in eq (13).
  unwrapping_likelihood_scale = 2.0f; //scale parameter for the unwrapping likelihood, s_1^2 in eq (15).
  phase_confidence_scale = 3.0f; //scale parameter for the phase likelihood, s_2^2 in eq (23)
  kde_threshold = 0.5f; //threshold on the KDE output in eq (25), defines the inlier/outlier rate trade-off

  kde_neigborhood_size = 5; //spatial support of the KDE, defines a filter size of (2*kde_neigborhood_size+1 x 2*kde_neigborhood_size+1)
  num_hyps = 2; //number of phase unwrapping hypothesis considered by the KDE in each pixel. Implemented values are 2 and 3.
  //a large kde_neigborhood_size improves performance but may remove fine structures and makes the processing slower.
  //setting num_hyp to 3 improves the performance slightly but makes processing slower