|
|
latest version v1.9 - last update 10 Apr 2010 |
|
Given a set of connected image regions (coded in a labeled mask or std::list of area point lists) and the original color image, this functor will compute a similarity measure between the objects, and will code them in a similarity matrix. More...
#include <ltiSimilarityMatrix.h>
Classes | |
| class | parameters |
| the parameters for the class similarityMatrix More... | |
Public Member Functions | |
| similarityMatrix () | |
| similarityMatrix (const similarityMatrix &other) | |
| virtual | ~similarityMatrix () |
| virtual const char * | getTypeName () const |
| bool | apply (const image &img, const int &nObj, const imatrix &lmask, dmatrix &colSimMat, dmatrix &geoSimMat) const |
| bool | apply (const image &img, const std::list< areaPoints > &lapts, dmatrix &colSimMat, dmatrix &geoSimMat, imatrix &lmask) const |
| bool | neighbors (const int &nObj, const imatrix &lmask, const int &label, ivector &neighborLabels) const |
| similarityMatrix & | copy (const similarityMatrix &other) |
| similarityMatrix & | operator= (const similarityMatrix &other) |
| virtual functor * | clone () const |
| const parameters & | getParameters () const |
Static Public Member Functions | |
Color Distances | |
| static double | euclideanLLL2 (const rgbPixel &a, const rgbPixel &b) |
| static double | euclideanRGB2 (const rgbPixel &a, const rgbPixel &b) |
| static double | chromaticity2 (const rgbPixel &a, const rgbPixel &b) |
| static double | hueSaturation (const rgbPixel &a, const rgbPixel &b) |
| static double | hue (const rgbPixel &a, const rgbPixel &b) |
| static double | neighbor (const rgbPixel &a, const rgbPixel &b) |
Protected Member Functions | |
| void | createMask (const std::list< areaPoints > &apts, imatrix &mask) const |
| void | similarity (const image &img, const imatrix &mask, dmatrix &colBorderSimMat, dmatrix &colAreaSimMat, dmatrix &geoSimMat) const |
Given a set of connected image regions (coded in a labeled mask or std::list of area point lists) and the original color image, this functor will compute a similarity measure between the objects, and will code them in a similarity matrix.
The first approach (BorderBased) will compute the sum of the square root of the difference of the colors at the borders between two both objects. The second approach (AreaBased) will compute the difference of the mean colors for each object.
See also Segmentation Overview
| lti::similarityMatrix::similarityMatrix | ( | ) |
default constructor
| lti::similarityMatrix::similarityMatrix | ( | const similarityMatrix & | other | ) |
copy constructor
| other | the object to be copied |
| virtual lti::similarityMatrix::~similarityMatrix | ( | ) | [virtual] |
destructor
| bool lti::similarityMatrix::apply | ( | const image & | img, | |
| const std::list< areaPoints > & | lapts, | |||
| dmatrix & | colSimMat, | |||
| dmatrix & | geoSimMat, | |||
| imatrix & | lmask | |||
| ) | const |
operates on the given parameter.
| img | original image. All points in the list of areaPoints must lie inside the image, or an exception will be thrown. | |
| lapts | list of area points. | |
| colSimMat | color similarity matrix | |
| geoSimMat | geometry similarity measure | |
| lmask | label mask computed from the area points |
| bool lti::similarityMatrix::apply | ( | const image & | img, | |
| const int & | nObj, | |||
| const imatrix & | lmask, | |||
| dmatrix & | colSimMat, | |||
| dmatrix & | geoSimMat | |||
| ) | const |
operates on the given parameter.
| img | original image. All points in the list of areaPoints must lie inside the image, or an exception will be thrown. | |
| nObj | number of objects in the mask (without background). Is expected here in order to reduce its computation from the mask. This value should be equal to the maximum value in the mask, i.e. the number of elements of the area points list used to generate the mask. | |
| lmask | label mask computed from the area points | |
| colSimMat | color similarity matrix | |
| geoSimMat | geometry similarity measure |
| static double lti::similarityMatrix::chromaticity2 | ( | const rgbPixel & | a, | |
| const rgbPixel & | b | |||
| ) | [static] |
euclidean distance in the rg color space, ignoring the intensity channel.
| virtual functor* lti::similarityMatrix::clone | ( | ) | const [virtual] |
returns a pointer to a clone of this functor.
Implements lti::functor.
| similarityMatrix& lti::similarityMatrix::copy | ( | const similarityMatrix & | other | ) |
copy data of "other" functor.
| other | the functor to be copied |
Reimplemented from lti::functor.
| void lti::similarityMatrix::createMask | ( | const std::list< areaPoints > & | apts, | |
| imatrix & | mask | |||
| ) | const [protected] |
create a mask from the area points.
The mask must have the correct dimensions before calling this method.
| static double lti::similarityMatrix::euclideanLLL2 | ( | const rgbPixel & | a, | |
| const rgbPixel & | b | |||
| ) | [static] |
This measure compares the differences between the color channels (red-green)^2, (red-blue)^2 and (green-blue)^2 for each pixel, and then the square of the differences between these values for a and b.
| static double lti::similarityMatrix::euclideanRGB2 | ( | const rgbPixel & | a, | |
| const rgbPixel & | b | |||
| ) | [static] |
square of the euclidean distance in RGB color space, i.e.
this returns a.distanceSqr(b)
| const parameters& lti::similarityMatrix::getParameters | ( | ) | const |
returns used parameters
Reimplemented from lti::functor.
| virtual const char* lti::similarityMatrix::getTypeName | ( | ) | const [virtual] |
returns the name of this type ("similarityMatrix")
Reimplemented from lti::functor.
distance of the pixels in the HSV color space, where only the hue channel is considered.
The angular distance of hue is always normed from 0 to 255.
Also usefull to ignore shadow based differences.
| static double lti::similarityMatrix::hueSaturation | ( | const rgbPixel & | a, | |
| const rgbPixel & | b | |||
| ) | [static] |
distance of the pixels in the HSV color space.
The value channel is ignored. The angular distance of hue is always normed from 0 to 255. The Saturation value is also normed.
Also usefull to ignore shadow based differences.
returns always 1.0 Usefull to create a matrix to know which areas are neighbors
| bool lti::similarityMatrix::neighbors | ( | const int & | nObj, | |
| const imatrix & | lmask, | |||
| const int & | label, | |||
| ivector & | neighborLabels | |||
| ) | const |
computes for the given mask which labels are used in neighbor objects for the given label.
| nObj | number of objects in the mask (without background). Is expected here in order to reduce its computation from the mask. This value should be equal to the maximum value in the mask, i.e. the number of elements of the area points list used to generate the mask. | |
| lmask | label mask computed from the area points | |
| label | label for which the neighbors need to be found | |
| neighborLabels | resulting vector |
| similarityMatrix& lti::similarityMatrix::operator= | ( | const similarityMatrix & | other | ) |
| void lti::similarityMatrix::similarity | ( | const image & | img, | |
| const imatrix & | mask, | |||
| dmatrix & | colBorderSimMat, | |||
| dmatrix & | colAreaSimMat, | |||
| dmatrix & | geoSimMat | |||
| ) | const [protected] |
create the border similarity matrix.
The size of the similarity matrix must be set before calling this method.
| img | original color image | |
| mask | labeled segmentation mask, created with createMask() | |
| colBorderSimMat | border color similarity mask, the size must be set previously | |
| colAreaSimMat | area color similarity | |
| geoSimMat | geomatrical similarity measure |