You can use this gaussian fit to get the correlation rho for any arbitrary distance.
The fit gaussian of the form f(h) = c exp( -(h/a)^2 ). c is called the sill and a is called the range. h is the distance between the two points.
Public Member Functions | |
| GaussianFit (const Correlogram &sample_rhos, size_t maxSamples=0) | |
| pass in the correlogram and how many samples you want to fit over. | |
| bool | isValid () const |
| won't be valid incase of singularity problems. | |
| float | getSill () const |
| Sill (c) of the fit equation. | |
| float | getRange () const |
| range (a) of the fit equation. | |
| float | getDistance (float rho) const |
| return the distance, h, at which the correlation falls below this value. | |
| float | getRho (float dist) const |
| get the correlation at a distance. | |
| code::Image< float > | computeKernel () const |
| code::Image< float > | computeQuarterKernel () const |
| w2img::Correlogram::GaussianFit::GaussianFit | ( | const Correlogram & | sample_rhos, | |
| size_t | maxSamples = 0 | |||
| ) |
pass in the correlogram and how many samples you want to fit over.
Zero indicates everything ...
| code::Image<float> w2img::Correlogram::GaussianFit::computeKernel | ( | ) | const |
| code::Image<float> w2img::Correlogram::GaussianFit::computeQuarterKernel | ( | ) | const |
| float w2img::Correlogram::GaussianFit::getDistance | ( | float | rho | ) | const |
return the distance, h, at which the correlation falls below this value.
| float w2img::Correlogram::GaussianFit::getRange | ( | ) | const [inline] |
range (a) of the fit equation.
You *can* pass in distances > range
| float w2img::Correlogram::GaussianFit::getRho | ( | float | dist | ) | const |
get the correlation at a distance.
| float w2img::Correlogram::GaussianFit::getSill | ( | ) | const [inline] |
Sill (c) of the fit equation.
| bool w2img::Correlogram::GaussianFit::isValid | ( | ) | const [inline] |
won't be valid incase of singularity problems.
1.4.7