pborsotti

Synopsis

Description

pborsotti computes a goodness measure for quantitative evaluation of gray levels, color and multispectral image segmentation results as defined by M. Borsotti^*.

The measure is defined from three criteria:

• regions must be uniform and homogeneous;
• the region's interiors must be simple, without too many small holes;
• adjacent regions must present significantly different values for uniform characteristics.

The measure is computed as follows:

F(I) = (1/(1000*A)) * sqrt(N) * sumR [ (ei2 / (1+log(Ai)) + (R(Ai)/ Ai)2)]

• A is the total area of regions.
• Ai is the area of the region i.
• N is the number of regions.
• R(Ai) is the number of regions that have the same area than Ai.
• ei is defined as the sum of euclidean distances between the color vectors of the pixel of the region i and the color vector attributed to region i.

The previous equation is composed of three terms:

1. a normalization factor that takes into account the size of the image;
2. a penalization factor for under-segmented regions;
3. the sum is composed of two parts:
   • penalization for small regions or heterogeneous of the regions;
   • penalization for region with the same size (over-segmentation).

The smaller the value of the Borsotti's measure is, the better the segmentation result should be.

Caution: Regions with label=0 are not considered for computing.

Inputs

• rg_in: a region map.
• im_in: an image.

Result

Returns a positive real value.
(Use pstatus to get this value).

Examples

Computes the borsotti measure for a simple binarization segmentation process:

   pbinarization 80 1e30 tangram.pan i1.pan
   plabeling 8 i1.pan i2.pan
   pborsotti i2.pan tangram.pan
   pstatus

See also

C++ prototype

Version française

Calcul du critère de qualité basé sur le nombre, l'aire et la variance des régions.

Reference

^*M. Borsotti, P. Campadelli, R. Schettini, "Quantitative evaluation of color image segmentation results", Pattern Recoginition Letters, 19:741-747, 1998.