本文整理汇总了C++中ImageData::GetChannelData方法的典型用法代码示例。如果您正苦于以下问题：C++ ImageData::GetChannelData方法的具体用法？C++ ImageData::GetChannelData怎么用？C++ ImageData::GetChannelData使用的例子？那么, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类ImageData的用法示例。
在下文中一共展示了ImageData::GetChannelData方法的1个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的C++代码示例。
示例1: Solve
ImageData IRLSMapSolver::Solve(const ImageData& initial_estimate) {
const int num_pixels = GetNumPixels();
const int num_channels = GetNumChannels();
const cv::Size image_size = GetImageSize();
CHECK_EQ(initial_estimate.GetNumPixels(), num_pixels);
CHECK_EQ(initial_estimate.GetNumChannels(), num_channels);
CHECK_EQ(initial_estimate.GetImageSize(), image_size);
// If the split_channels option is set, loop over the channels here and solve
// them independently. Otherwise, solve all channels at once.
const int num_channels_per_split =
solver_options_.split_channels ? 1 : num_channels;
const int num_solver_rounds = num_channels / num_channels_per_split;
const int num_data_points = num_channels_per_split * num_pixels;
if (num_channels_per_split != num_channels) {
LOG(INFO) << "Splitting up image into " << num_solver_rounds
<< " sections with " << num_channels_per_split
<< " channel(s) in each section.";
}
// Scale the option stop criteria parameters based on the number of
// parameters and strength of the regularizers.
IRLSMapSolverOptions solver_options_scaled = solver_options_;
solver_options_scaled.AdjustThresholdsAdaptively(
num_data_points, GetRegularizationParameterSum());
if (IsVerbose()) {
solver_options_scaled.PrintSolverOptions();
}
ImageData estimated_image;
for (int i = 0; i < num_solver_rounds; ++i) {
if (num_solver_rounds > 1) {
LOG(INFO) << "Starting solver on image subset #" << (i + 1) << ".";
}
const int channel_start = i * num_channels_per_split;
const int channel_end = channel_start + num_channels_per_split;
// Copy the initial estimate data (within the appropriate channel range) to
// the solver's array.
alglib::real_1d_array solver_data;
solver_data.setlength(num_data_points);
for (int channel = 0; channel < num_channels_per_split; ++channel) {
double* data_ptr = solver_data.getcontent() + (num_pixels * channel);
const double* channel_ptr = initial_estimate.GetChannelData(
channel_start + channel);
std::copy(channel_ptr, channel_ptr + num_pixels, data_ptr);
}
// Set up the base objective function (just data term). The regularization
// term depends on the IRLS weights, so it gets added in the IRLS loop.
ObjectiveFunction objective_function_data_term_only(num_data_points);
std::shared_ptr<ObjectiveTerm> data_term(new ObjectiveDataTerm(
image_model_, observations_, channel_start, channel_end, image_size));
objective_function_data_term_only.AddTerm(data_term);
RunIRLSLoop(
solver_options_scaled,
objective_function_data_term_only,
regularizers_,
image_size,
channel_start,
channel_end,
&solver_data);
for (int channel = 0; channel < num_channels_per_split; ++channel) {
const double* data_ptr =
solver_data.getcontent() + (num_pixels * channel);
estimated_image.AddChannel(data_ptr, image_size);
}
}
return estimated_image;
}