Point Cloud Library (PCL)  1.14.1-dev
fast_bilateral_omp.hpp
1 /*
2  * Software License Agreement (BSD License)
3  *
4  * Point Cloud Library (PCL) - www.pointclouds.org
5  * Copyright (c) 2012-, Open Perception, Inc.
6  * Copyright (c) 2004, Sylvain Paris and Francois Sillion
7 
8  * All rights reserved.
9 
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  *
14  * * Redistributions of source code must retain the above copyright
15  * notice, this list of conditions and the following disclaimer.
16  * * Redistributions in binary form must reproduce the above
17  * copyright notice, this list of conditions and the following
18  * disclaimer in the documentation and/or other materials provided
19  * with the distribution.
20  * * Neither the name of the copyright holder(s) nor the names of its
21  * contributors may be used to endorse or promote products derived
22  * from this software without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25  * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
27  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
28  * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
29  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
30  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
31  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
32  * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
34  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35  * POSSIBILITY OF SUCH DAMAGE.
36  *
37  * $Id: fast_bilateral_omp.hpp 8381 2013-01-02 23:12:44Z sdmiller $
38  *
39  */
40 #ifndef PCL_FILTERS_IMPL_FAST_BILATERAL_OMP_HPP_
41 #define PCL_FILTERS_IMPL_FAST_BILATERAL_OMP_HPP_
42 
43 #include <pcl/filters/fast_bilateral_omp.h>
44 #include <pcl/common/io.h>
45 
46 //////////////////////////////////////////////////////////////////////////////////////////////
47 template <typename PointT> void
49 {
50  if (nr_threads == 0)
51 #ifdef _OPENMP
52  threads_ = omp_get_num_procs();
53 #else
54  threads_ = 1;
55 #endif
56  else
57  threads_ = nr_threads;
58 }
59 
60 //////////////////////////////////////////////////////////////////////////////////////////////
61 template <typename PointT> void
63 {
64  if (!input_->isOrganized ())
65  {
66  PCL_ERROR ("[pcl::FastBilateralFilterOMP] Input cloud needs to be organized.\n");
67  return;
68  }
69 
70  copyPointCloud (*input_, output);
71  float base_max = -std::numeric_limits<float>::max (),
72  base_min = std::numeric_limits<float>::max ();
73  bool found_finite = false;
74  for (const auto& pt: output)
75  {
76  if (std::isfinite(pt.z))
77  {
78  base_max = std::max<float>(pt.z, base_max);
79  base_min = std::min<float>(pt.z, base_min);
80  found_finite = true;
81  }
82  }
83  if (!found_finite)
84  {
85  PCL_WARN ("[pcl::FastBilateralFilterOMP] Given an empty cloud. Doing nothing.\n");
86  return;
87  }
88 #pragma omp parallel for \
89  default(none) \
90  shared(base_min, base_max, output) \
91  num_threads(threads_)
92  for (long int i = 0; i < static_cast<long int> (output.size ()); ++i)
93  if (!std::isfinite (output.at(i).z))
94  output.at(i).z = base_max;
95 
96  const float base_delta = base_max - base_min;
97 
98  const std::size_t padding_xy = 2;
99  const std::size_t padding_z = 2;
100 
101  const std::size_t small_width = static_cast<std::size_t> (static_cast<float> (input_->width - 1) / sigma_s_) + 1 + 2 * padding_xy;
102  const std::size_t small_height = static_cast<std::size_t> (static_cast<float> (input_->height - 1) / sigma_s_) + 1 + 2 * padding_xy;
103  const std::size_t small_depth = static_cast<std::size_t> (base_delta / sigma_r_) + 1 + 2 * padding_z;
104 
105  Array3D data (small_width, small_height, small_depth);
106 #if OPENMP_LEGACY_CONST_DATA_SHARING_RULE
107 #pragma omp parallel for \
108  default(none) \
109  shared(base_min, data, output) \
110  num_threads(threads_)
111 #else
112 #pragma omp parallel for \
113  default(none) \
114  shared(base_min, data, output, small_height, small_width) \
115  num_threads(threads_)
116 #endif
117  for (long int i = 0; i < static_cast<long int> (small_width * small_height); ++i)
118  {
119  auto small_x = static_cast<std::size_t> (i % small_width);
120  auto small_y = static_cast<std::size_t> (i / small_width);
121  auto start_x = static_cast<std::size_t>(
122  std::max ((static_cast<float> (small_x) - static_cast<float> (padding_xy) - 0.5f) * sigma_s_ + 1, 0.f));
123  auto end_x = static_cast<std::size_t>(
124  std::max ((static_cast<float> (small_x) - static_cast<float> (padding_xy) + 0.5f) * sigma_s_ + 1, 0.f));
125  auto start_y = static_cast<std::size_t>(
126  std::max ((static_cast<float> (small_y) - static_cast<float> (padding_xy) - 0.5f) * sigma_s_ + 1, 0.f));
127  auto end_y = static_cast<std::size_t>(
128  std::max ((static_cast<float> (small_y) - static_cast<float> (padding_xy) + 0.5f) * sigma_s_ + 1, 0.f));
129  for (std::size_t x = start_x; x < end_x && x < input_->width; ++x)
130  {
131  for (std::size_t y = start_y; y < end_y && y < input_->height; ++y)
132  {
133  const float z = output (x,y).z - base_min;
134  const std::size_t small_z = static_cast<std::size_t> (static_cast<float> (z) / sigma_r_ + 0.5f) + padding_z;
135  Eigen::Vector2f& d = data (small_x, small_y, small_z);
136  d[0] += output (x,y).z;
137  d[1] += 1.0f;
138  }
139  }
140  }
141 
142  std::vector<long int> offset (3);
143  offset[0] = &(data (1,0,0)) - &(data (0,0,0));
144  offset[1] = &(data (0,1,0)) - &(data (0,0,0));
145  offset[2] = &(data (0,0,1)) - &(data (0,0,0));
146 
147  Array3D buffer (small_width, small_height, small_depth);
148 
149  for (std::size_t dim = 0; dim < 3; ++dim)
150  {
151  for (std::size_t n_iter = 0; n_iter < 2; ++n_iter)
152  {
153  Array3D* current_buffer = (n_iter % 2 == 1 ? &buffer : &data);
154  Array3D* current_data =(n_iter % 2 == 1 ? &data : &buffer);
155 #if OPENMP_LEGACY_CONST_DATA_SHARING_RULE
156 #pragma omp parallel for \
157  default(none) \
158  shared(current_buffer, current_data, dim, offset) \
159  num_threads(threads_)
160 #else
161 #pragma omp parallel for \
162  default(none) \
163  shared(current_buffer, current_data, dim, offset, small_depth, small_height, small_width) \
164  num_threads(threads_)
165 #endif
166  for(long int i = 0; i < static_cast<long int> ((small_width - 2)*(small_height - 2)); ++i)
167  {
168  auto x = static_cast<std::size_t> (i % (small_width - 2) + 1);
169  auto y = static_cast<std::size_t> (i / (small_width - 2) + 1);
170  const long int off = offset[dim];
171  Eigen::Vector2f* d_ptr = &(current_data->operator() (x,y,1));
172  Eigen::Vector2f* b_ptr = &(current_buffer->operator() (x,y,1));
173 
174  for(std::size_t z = 1; z < small_depth - 1; ++z, ++d_ptr, ++b_ptr)
175  *d_ptr = (*(b_ptr - off) + *(b_ptr + off) + 2.0 * (*b_ptr)) / 4.0;
176  }
177  }
178  }
179  // Note: this works because there are an even number of iterations.
180  // If there were an odd number, we would need to end with a:
181  // std::swap (data, buffer);
182 
183  if (early_division_)
184  {
185  for (auto d = data.begin (); d != data.end (); ++d)
186  *d /= ((*d)[0] != 0) ? (*d)[1] : 1;
187 
188 #pragma omp parallel for \
189  default(none) \
190  shared(base_min, data, output) \
191  num_threads(threads_)
192  for (long int i = 0; i < static_cast<long int> (input_->size ()); ++i)
193  {
194  auto x = static_cast<std::size_t> (i % input_->width);
195  auto y = static_cast<std::size_t> (i / input_->width);
196  const float z = output (x,y).z - base_min;
197  const Eigen::Vector2f D = data.trilinear_interpolation (static_cast<float> (x) / sigma_s_ + padding_xy,
198  static_cast<float> (y) / sigma_s_ + padding_xy,
199  z / sigma_r_ + padding_z);
200  output(x,y).z = D[0];
201  }
202  }
203  else
204  {
205 #pragma omp parallel for \
206  default(none) \
207  shared(base_min, data, output) \
208  num_threads(threads_)
209  for (long i = 0; i < static_cast<long int> (input_->size ()); ++i)
210  {
211  auto x = static_cast<std::size_t> (i % input_->width);
212  auto y = static_cast<std::size_t> (i / input_->width);
213  const float z = output (x,y).z - base_min;
214  const Eigen::Vector2f D = data.trilinear_interpolation (static_cast<float> (x) / sigma_s_ + padding_xy,
215  static_cast<float> (y) / sigma_s_ + padding_xy,
216  z / sigma_r_ + padding_z);
217  output (x,y).z = D[0] / D[1];
218  }
219  }
220 }
221 
222 
223 
224 #endif /* PCL_FILTERS_IMPL_FAST_BILATERAL_OMP_HPP_ */
225 
void setNumberOfThreads(unsigned int nr_threads=0)
Initialize the scheduler and set the number of threads to use.
void applyFilter(PointCloud &output) override
Filter the input data and store the results into output.
typename FastBilateralFilter< PointT >::Array3D Array3D
PointCloud represents the base class in PCL for storing collections of 3D points.
Definition: point_cloud.h:173
const PointT & at(int column, int row) const
Obtain the point given by the (column, row) coordinates.
Definition: point_cloud.h:262
std::size_t size() const
Definition: point_cloud.h:443
void copyPointCloud(const pcl::PointCloud< PointInT > &cloud_in, pcl::PointCloud< PointOutT > &cloud_out)
Copy all the fields from a given point cloud into a new point cloud.
Definition: io.hpp:142