void run ()
{
Header H_in = Header::open (argument[0]);
Sparse::Image<FixelMetric> fixel (H_in);
const size_t lmax = 8;
const size_t N = Math::SH::NforL (lmax);
Math::SH::aPSF<default_type> aPSF (lmax);
Header H_out (H_in);
H_out.datatype() = DataType::Float32;
H_out.datatype().set_byte_order_native();
const size_t sh_dim = H_in.ndim();
H_out.ndim() = H_in.ndim() + 1;
H_out.size (sh_dim) = N;
auto sh = Image<float>::create (argument[1], H_out);
std::vector<default_type> values;
Eigen::Matrix<default_type, Eigen::Dynamic, 1> apsf_values;
for (auto l1 = Loop("converting sparse fixel data to SH image", fixel) (fixel, sh); l1; ++l1) {
values.assign (N, 0.0);
for (size_t index = 0; index != fixel.value().size(); ++index) {
apsf_values = aPSF (apsf_values, fixel.value()[index].dir);
const default_type scale_factor = fixel.value()[index].value;
for (size_t i = 0; i != N; ++i)
values[i] += apsf_values[i] * scale_factor;
}
for (auto l2 = Loop(sh_dim) (sh); l2; ++l2)
sh.value() = values[sh.index(sh_dim)];
}
}
void run ()
{
Header H_in = Header::open (argument[0]);
Sparse::Image<FixelMetric> in (H_in);
const int op = argument[1];
Header H_out (H_in);
H_out.datatype() = DataType::Float32;
H_out.datatype().set_byte_order_native();
H_out.keyval().erase (Sparse::name_key);
H_out.keyval().erase (Sparse::size_key);
if (op == 10) { // count
H_out.datatype() = DataType::UInt8;
} else if (op == 13 || op == 14) { // dec
H_out.set_ndim (4);
H_out.size (3) = 3;
} else if (op == 15 || op == 16 || op == 17) { // split_*
H_out.set_ndim (4);
uint32_t max_count = 0;
for (auto l = Loop ("determining largest fixel count", in) (in); l; ++l)
max_count = std::max (max_count, in.value().size());
if (max_count == 0)
throw Exception ("fixel image is empty");
// 3 volumes per fixel if performing split_dir
H_out.size(3) = (op == 17) ? (3 * max_count) : max_count;
}
auto out = Image<float>::create (argument[2], H_out);
auto opt = get_options ("weighted");
const bool weighted = opt.size();
auto loop = ThreadedLoop ("converting sparse fixel data to scalar image", in);
switch (op) {
case 0: loop.run (Mean (weighted), in, out); break;
case 1: loop.run (Sum (weighted), in, out); break;
case 2: loop.run (Product (weighted), in, out); break;
case 3: loop.run (RMS (weighted), in, out); break;
case 4: loop.run (Var (weighted), in, out); break;
case 5: loop.run (Std (weighted), in, out); break;
case 6: loop.run (Min (weighted), in, out); break;
case 7: loop.run (Max (weighted), in, out); break;
case 8: loop.run (AbsMax (weighted), in, out); break;
case 9: loop.run (MagMax (weighted), in, out); break;
case 10: loop.run (Count (weighted), in, out); break;
case 11: loop.run (Complexity (weighted), in, out); break;
case 12: loop.run (SF (weighted), in, out); break;
case 13: loop.run (DEC_unit (weighted), in, out); break;
case 14: loop.run (DEC_scaled (weighted), in, out); break;
case 15: loop.run (SplitSize (weighted), in, out); break;
case 16: loop.run (SplitValue (weighted), in, out); break;
case 17: loop.run (SplitDir (weighted), in, out); break;
}
}
void run ()
{
auto header = Header::open (argument[0]);
Sparse::Image<FixelMetric> input (argument[0]);
Sparse::Image<FixelMetric> output (argument[1], header);
for (auto i = Loop ("computing log", input) (input, output); i; ++i) {
output.value().set_size (input.value().size());
for (size_t fixel = 0; fixel != input.value().size(); ++fixel) {
output.value()[fixel] = input.value()[fixel];
output.value()[fixel].value = std::log (input.value()[fixel].value);
}
}
}