Eigen::MatrixXd sampleRandomPatches(const Eigen::MatrixXd& images,
int channels, int rows, int cols,
int samples, int patchRows, int patchCols)
{
RandomNumberGenerator rng;
const int channelSize = rows * cols;
const int patchSize = channels * patchRows * patchCols;
const int totalPatches = images.rows() * samples;
Eigen::MatrixXd patches(totalPatches, patchSize);
std::vector<std::pair<int, int> > patchIndices;
patchIndices.reserve(samples);
for(int n = 0; n < totalPatches; n++)
patchIndices.push_back(std::make_pair<int>(
rng.generateInt(0, rows - patchRows + 1),
rng.generateInt(0, cols - patchCols + 1)));
#pragma omp parallel for
for(int m = 0; m < images.rows(); ++m)
{
for(int n = 0; n < samples; ++n)
{
const int patchIdx = m*samples+n;
const int rowStart = patchIndices[patchIdx].first;
const int colStart = patchIndices[patchIdx].second;
for(int chan = 0, channelOffset = 0, pxIdx = 0; chan < channels;
++chan, channelOffset += channelSize)
{
for(int row = 0; row < patchRows; ++row)
for(int col = 0; col < patchCols; ++col)
patches(patchIdx, pxIdx++) =
images(m, channelOffset+(rowStart+row)*cols+colStart+col);
}
}
}
return patches;
}
