pmeanshiftsegmentation

Synopsis

Description

The pmeanshiftsegmentation classifies the input image pixels into regions using color homogeneity. It is a procedure for locating the maxima of a density function given discrete data sampled from that function. The algorithm is based on the detection of the modes of this density function.

The parameters controls the classification process. The spatial-bandwidth controls the size of the search window when computing mean shift. The range-bandwidth controls the color vicinity for homogeneous regions. The more the spatial-bandwidth and the range-bandwith, the less number of regions.

The result is the region map rg_out.

Parameters

• spatial-bandwidth specifies a spatial search window of size (2r+1)*(2r+1) during the mean shift computation, where r is the spatial bandwidth. It's an integer value greater than zero.
• range-bandwidth specifies the bandwidth of the search window in the range subspace during the computation of mean shift. It's a real value greater than zero (should not be high, generally <10).
• minimum-region-area specifies the minimum allowable region area (in pixels) contained in the segmented image. It's an integer greater than zero given in pixels.
• accuracy-level determines the accuracy level used when computing mean shift. Value is an integer in [0..2], where 2 is the maximum accuracy level. However the more is the accuracy level, the less is speed level. 2 gives the most accurate results. Note that the size of the spatial-bandwidth parameter also infers on the speed.

Inputs

• im_in: a 2D image.

Outputs

• rg_out: a region map.

Result

Returns the number of regions or FAILURE.

Examples

Segments the tangram pieces:

  pmeanshiftsegmentation 7 6.5 100 0 examples/tangram.pan a.pan

See also

C++ prototype

Reference