static cell_regression_stats *get_cell_regression_stats(
grid<pixel_t> const &red, grid<pixel_t> const &nir, size_t size,
bool force)
{
size_t output_block = 100;
rect<size_t> sub = {0, 0, size, size};
size_t xCount = red.width() / size;
size_t yCount = red.height() / size;
IF_NORMAL(std::cout << yCount << " rows" << std::endl);
size_t cell_count = xCount * yCount;
array<numeric_t> red_avg(cell_count);
array<numeric_t> nir_avg(cell_count);
array<bool> goodness(cell_count);
size_t good_count = 0;
bool *goodness_ptr = goodness.data();
numeric_t *red_avg_ptr = red_avg.data();
numeric_t *nir_avg_ptr = nir_avg.data();
for (sub.y = 0; sub.y + size <= red.height(); sub.y += size) {
for (sub.x = 0; sub.x + size <= red.width();
sub.x += size) {
const grid<pixel_t> sub_red(
const_cast<grid<pixel_t>*>(&red), sub);
const grid<pixel_t> sub_nir(
const_cast<grid<pixel_t>*>(&nir), sub);
if (force || is_good_data(sub_red, sub_nir)) {
*red_avg_ptr = stats_mean(sub_red);
*nir_avg_ptr = stats_mean(sub_nir);
*goodness_ptr = true;
good_count++;
} else {
IF_VERBOSE(std::cout << "Bad sector: ("
);
IF_VERBOSE(std::cout << sub.x << ", ");
IF_VERBOSE(std::cout << sub.y << ") ");
*goodness_ptr = false;
}
red_avg_ptr++;
nir_avg_ptr++;
goodness_ptr++;
}
if ((sub.y / size) % output_block == 0) {
IF_NORMAL(std::cout << "\ranalyzing rows ");
IF_NORMAL(std::cout << "starting at ");
IF_NORMAL(std::cout << (sub.y / size));
IF_NORMAL(std::cout << "...");
IF_NORMAL(std::cout.flush());
}
}
IF_NORMAL(std::cout << std::endl);
IF_VERBOSE(std::cout << "Found " << good_count << " good ");
IF_VERBOSE(std::cout << "sectors out of " << cell_count);
IF_VERBOSE(std::cout << std::endl);
array<numeric_t> good_red_data(good_count);
array<numeric_t> good_nir_data(good_count);
size_t good_data_cur = 0;
for (size_t i = 0; i < cell_count; i++) {
if (goodness[i]) {
good_red_data[good_data_cur] = red_avg[i];
good_nir_data[good_data_cur] = nir_avg[i];
good_data_cur++;
}
}
linear_regression *reg = numeric_find_linear_regression(
good_red_data, good_nir_data);
cell_regression_stats *stats = new cell_regression_stats;
stats->cell_size = size;
stats->regression = *reg;
delete reg;
return stats;
}
long double stats_summarize(stats *stats) {
qsort(stats->data, stats->limit, sizeof(uint64_t), &stats_compare);
return stats_mean(stats);
}
