org.openimaj.feature.local.matcher

Class KeypointCorrespondenceTestHelper

java.lang.Object

org.openimaj.feature.local.matcher.KeypointCorrespondenceTestHelper

public class KeypointCorrespondenceTestHelper
extends Object

Supports the generation of mild affine transforms. This is especially useful for the checking of localfeatures which should show some resiliance to such transforms

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk), Sina Samangooei (ss@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
KeypointCorrespondenceTestHelper()

Method Summary

Modifier and Type	Method and Description
static float	correspondance(List<Pair<Keypoint>> matches, Jama.Matrix transform) Provide simplistic check of correspondence between the keypoints extracted and a given transform.
static float	correspondance(List<Pair<Keypoint>> matches, Jama.Matrix transform, float scaleThreshold) Provide simplistic check of correspondence between the keypoints extracted and a given transform.
static Jama.Matrix	generateMildTransform(FImage image) Generate an affine transform between some reasonable limits: - Rotate it between 0 and 360 - transform it between 0 and 10 pixels - slant it (sheer in x and y) between 0 and 10 degrees - scale it between 0 and 2.0 times
static Jama.Matrix	generateMildTransform(FImage image, Random seed) Generate an affine transform between some reasonable limits: - Rotate it between 0 and 360 - transform it between 0 and 10 pixels - slant it (sheer in x and y) between 0 and 10 degrees - scale it between 0 and 2.0 times

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

KeypointCorrespondenceTestHelper

public KeypointCorrespondenceTestHelper()

Method Detail

generateMildTransform

public static Jama.Matrix generateMildTransform(FImage image)

Generate an affine transform between some reasonable limits: - Rotate it between 0 and 360 - transform it between 0 and 10 pixels - slant it (sheer in x and y) between 0 and 10 degrees - scale it between 0 and 2.0 times

Parameters:

image - the image on which to generate the transform

Returns:

a 3x3 transformation matrix

generateMildTransform

public static Jama.Matrix generateMildTransform(FImage image,
                                                Random seed)

Generate an affine transform between some reasonable limits: - Rotate it between 0 and 360 - transform it between 0 and 10 pixels - slant it (sheer in x and y) between 0 and 10 degrees - scale it between 0 and 2.0 times

Parameters:

image - the image on which to generate the transform

seed - the random seed against which to generate the random mild transforms

Returns:

a 3x3 transformation matrix

correspondance

public static float correspondance(List<Pair<Keypoint>> matches,
                                   Jama.Matrix transform)

Provide simplistic check of correspondence between the keypoints extracted and a given transform. The euclidian distance between the expected position of a keypoint and it's actual position. Note this should not be used for consistency matching, please see the LocalFeatureMatcher for better implementations for that purpose.

Parameters:

matches - list of keypoint matches

transform - the supposed transform applied

Returns:

lower numbers are lower euclidian distances

correspondance

public static float correspondance(List<Pair<Keypoint>> matches,
                                   Jama.Matrix transform,
                                   float scaleThreshold)

Provide simplistic check of correspondence between the keypoints extracted and a given transform. The euclidian distance between the expected position of a keypoint and it's actual position. Note this should not be used for consistency matching, please see the LocalFeatureMatcher for better implementations for that purpose.

Parameters:

matches - list of keypoint matches

transform - the supposed transform applied

scaleThreshold - The allowable scale shift between two keypoints

Returns:

lower numbers are lower euclidian distances