org.openimaj.math.geometry.transforms.estimation.sampling

Class BucketingSampler2d

java.lang.Object

org.openimaj.math.geometry.transforms.estimation.sampling.BucketingSampler2d

All Implemented Interfaces:

CollectionSampler<IndependentPair<Point2d,Point2d>>

@Reference(type=Article,
           author={"Zhengyou Zhang","Rachid Deriche","Olivier Faugeras","Quang-Tuan Luong"},
           title="A robust technique for matching two uncalibrated images through the recovery of the unknown epipolar geometry ",
           year="1995",
           journal="Artificial Intelligence ",
           pages={"87 "," 119"},
           url="http://www.sciencedirect.com/science/article/pii/0004370295000224",
           note="Special Volume on Computer Vision ",
           number="1--2",
           volume="78",
           customData={"issn","0004-3702","doi","http://dx.doi.org/10.1016/0004-3702(95)00022-4","keywords","Correlation "})
public class BucketingSampler2d
extends Object
implements CollectionSampler<IndependentPair<Point2d,Point2d>>

Implementation of the bucketing sampling strategy proposed by Zhang et al to try and ensure a good spatial distribution of point-pairs for estimation of geometric transforms and the fundamental matrix.

Works by splitting the space of first image points into a number of buckets. When selecting the sample, a bucket is chosen at random, with the probability of the bucket being picked weighted towards buckets with more points. Then a point pair is picked using a uniform random from within the bucket. Additionally, the algorithm attempts to pick unique buckets for each point by attempting to discount previously selected buckets if possible.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Field Summary

Fields

Modifier and Type	Field and Description
static int	DEFAULT_N_BUCKETS_PER_DIM Default number of buckets per dimension
static int	NUM_TRIALS Maximum allowed number of trials in picking a bucket that has not been previously picked.

Constructor Summary

Constructors

Constructor and Description
BucketingSampler2d() Construct the sampler with the default number of buckets in the x and y dimensions (8).
BucketingSampler2d(int nBucketsX, int nBucketsY) Construct the sampler with the given number of buckets in each dimension.

Method Summary

Modifier and Type	Method and Description
List<IndependentPair<Point2d,Point2d>>	sample(int nItems) Samples nItems items from the collection set by CollectionSampler.setCollection(Collection), returning a new collection with the given samples.
void	setCollection(Collection<? extends IndependentPair<Point2d,Point2d>> collection) Set the items to sample from.

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_N_BUCKETS_PER_DIM

public static final int DEFAULT_N_BUCKETS_PER_DIM

Default number of buckets per dimension

See Also:

Constant Field Values

NUM_TRIALS

public static int NUM_TRIALS

Maximum allowed number of trials in picking a bucket that has not been previously picked.

Constructor Detail

BucketingSampler2d

public BucketingSampler2d()

Construct the sampler with the default number of buckets in the x and y dimensions (8).

BucketingSampler2d

public BucketingSampler2d(int nBucketsX,
                          int nBucketsY)

Construct the sampler with the given number of buckets in each dimension.

Parameters:

nBucketsX - number of buckets in the x dimension

nBucketsY - number of buckets in the y dimension

Method Detail

setCollection

public void setCollection(Collection<? extends IndependentPair<Point2d,Point2d>> collection)

Description copied from interface: CollectionSampler

Set the items to sample from. This should only be called once for any given set of items. Implementors can assume that calling this resets any state of the sampler.

Specified by:

setCollection in interface CollectionSampler<IndependentPair<Point2d,Point2d>>

Parameters:

collection - the items to sample from

sample

public List<IndependentPair<Point2d,Point2d>> sample(int nItems)

Description copied from interface: CollectionSampler

Samples nItems items from the collection set by CollectionSampler.setCollection(Collection), returning a new collection with the given samples. Implementations can decide what to do if nItems is bigger than the number of items in the collection.

Specified by:

sample in interface CollectionSampler<IndependentPair<Point2d,Point2d>>

Parameters:

nItems - the number of items to sample

Returns:

the sample; can be null if a sample cannot be made for any reason.