pvarianceselection

Selects regions from variance factor.

Synopsis

pvarianceselection relation threshold [-m mask] [rg_in|-] [im_in|-] [rg_out|-]

Description

pvarianceselection selects regions from their variance factor. The parameter relation specifies the order relation to the threshold value that is used to select or not a region.

The region variance is calculated as follows:

   var = ((n*sigma2)-(sigma*sigma)) / (n*n)

where sigma is the sum of gray levels of the region,
where sigma2 is the sum of the square of the gray levels of the region,
where n is the number of pixels of the region.

Parameters

relation is an integer from [-3,3] which specifies the order relation to the threshold value:
- relation = 3: regions with the maximum value.
- relation = 2: regions >= threshold.
- relation = 1: regions > threshold.
- relation = 0: regions = threshold.
- relation = -1: regions < threshold.
- relation = -2: regions <= threshold.
- relation = -3: regions with the minimum value.
threshold is a real positive value.

Inputs

rg_in: a region map.
im_in: a grayscale image.

Outputs

rg_out: a region map.

Result

Returns the number of selected regions.

Examples

Selects regions with a variance ≤ 20:

   pvarianceselection -2 20 rin.pan rout.pan

C++ prototype

Errc PVarianceSelection( const Reg2d &rg_in, Img2duc &im_in, Reg2d &rg_out, int relation, float threshold );

Version française

Sélection de régions sur leur valeur de variance.

Author: Régis Clouard