main.m

warning off;
dbstop if error;
rand('seed', 255);
clear;
clc;

%% prepare save folders
config_dir = './config';
if ~exist(config_dir, 'dir')
  mkdir(config_dir);
end
save_dir = './results';
log_file = fullfile(save_dir, 'log.txt');
if ~exist(save_dir, 'dir')
  mkdir(save_dir);
end

%% denoise models
model_names = ["anchor", "daratech", "dc", "gargoyle", "lordquas"];
for i = 1 : 5
  model_name = char(model_names(i));
  gt_name = ['./models/gt/', model_name, '_GT.ply'];
  noise_level = num2str(0.05);
  noise_name = ['./models/noise/', model_name, '_gaussian_noise_', noise_level, '.ply'];

  % check if the parameter configuration file exists
  config_file = [model_name, '_', noise_level, '.mat'];
  config_file = fullfile(config_dir, config_file);
  if ~exist(config_file, 'file')
    kMaxIteration = 30;
    kDownSampleRate = 0.3;
    kPatchSize = 9;
    kPatchNeighborCount = 10;
    kChangeTolerance = 5e-5;

    alpha = 25.0 * (exp((1:kMaxIteration) / 20) - 1);
    alpha = repmat(alpha, 5, 1);
    alpha = reshape(alpha, 5 * kMaxIteration, 1);

    use_matlab_down_sample = false;
    use_normal = true;
    use_cosine = false;
    use_color = false;
    C = 5;
    save_iter_cloud = false;

    save(config_file, ...
      'kMaxIteration', ...
      'kDownSampleRate', ...
      'kPatchSize', ...
      'kPatchNeighborCount', ...
      'kChangeTolerance', ...
      'alpha', ...
      'use_matlab_down_sample', ...
      'use_normal', ...
      'use_cosine', ...
      'use_color', ...
      'C', ...
      'save_iter_cloud');
  else
    load(config_file);
  end

  gt_cloud = pcread(gt_name);
  noisy_cloud = pcread(noise_name);
  [best_mse, best_snr, max_iterations] = denoise_with_learning( ...
    noisy_cloud, gt_cloud, model_name, ...
    kMaxIteration, kChangeTolerance, ...
    kDownSampleRate, kPatchSize, kPatchNeighborCount, ...
    alpha, ...
    use_matlab_down_sample, ...
    use_normal, ...
    use_cosine, ...
    use_color, ...
    C, ...
    save_iter_cloud, ...
    log_file, ...
    save_dir);
end