CN101567051A - Image matching method based on characteristic points - Google Patents

Abstract

The invention belongs to the technical fields of mode indentifying and image processing, and relates to an image matching method based on characteristic points. The image matching method based on characteristic points defines an annular corner code model according to the K neighboring structure of the characteristic points and applies the model to point mode matching. Similar annular corner codes not only describe the similarity of the space structures of two characteristic points, but also can be used for local mapping estimation. The similarity of the two characteristic points is confirmed by the similar length of the largest similar annular corner codes associated with the two characteristic points. The image matching method based on characteristic points carries out structure matching according to the similarity of the local space structures of the characteristic points, and uses the local mapping and clustering to carry out optimization matching. The optimization matching result is used for optimizing the mapping estimation. The method can be used for the image matching based on the characteristic points, image jointing and embedding, motion tracing, information fusion of medical single-mode and multi-mode images, as well as image searching based on contents.

Description

A kind of method for registering images based on unique point

Technical field

The invention belongs to pattern-recognition and technical field of image processing, be specifically related to a kind of method for registering images based on unique point.This method can be used for the image registration based on unique point, image mosaic with inlay motion tracking, the information fusion of medical science single mode and multi-modality images, and CBIR etc.

Background technology

Point Pattern Matching can be understood as the images match based on feature, estimates from two points to be matched concentrated search corresponding relation and to the mapping between the point set.Point Pattern Matching is widely used in remote sensing (image mosaic, image mosaic), medical image (state of an illness diagnosis and tracking) and computer vision (target or scene Recognition, motion tracking).

Though a lot of in recent years outstanding Point Pattern Matching algorithms are suggested, almost there is not a kind of algorithm on computing time and precision, to satisfy application requirements simultaneously.Present disclosed algorithm is broadly divided into 5 classes: clustering algorithm (document that sees reference [1] [2]), shape contextual algorithms (document that sees reference [3]), loose labeling algorithm (document that sees reference [4] [5]), optimized Algorithm (document that sees reference [6]) progressively, figure matching algorithm (document that sees reference [7]).

Wherein,

Clustering algorithm hypothesis transformation model is similarity transformation, only need two pairs of match points because estimate similarity transformation, algorithm is at first concentrated from two points to be matched and is selected two pairs of points to make up, and estimates conversion coefficient then, verifies that with the conversion of estimating other points that may mate are right.Because the point with correct coupling also is correct to the conversion of estimating, carry out cluster with this conversion, then can institute might correct match point poly-ly arrive such, and if the point that correctly mates many, class also can be very big accordingly; On the contrary, if with the conversion to estimating of the point of erroneous matching, then this conversion is wrong certainly, and it is exactly randomness that erroneous conversions has characteristics.Erroneous matching that is to say that making sorter with erroneous conversions carries out cluster, then only converges to seldom point mostly at random.Algorithm carries out cluster according to this character to two feature point sets, and in the cluster result, maximum kind will be as corresponding relation output to be used for subsequent treatment.The defective of this class algorithm is the computation complexity height, and is consuming time, and if two feature point set difference in size very big, algorithm performance also can non-constant.Hough transformation was used to pre-service afterwards with speed-up computation, yet coupling remains a kind of work very consuming time, and was and very many at noise spot, and perhaps under two very large situations of point set difference in size, algorithm not necessarily guarantees to find out best correspondence.

The shape contextual algorithms also is to study the Point Pattern Matching technology of comparative heat in the last few years.The shape context-descriptive point concentrate other unique points space distribution information of certain unique point relatively.The unique point of mating in the same shape of two width of cloth images has the similar shapes context, each unique point correspondence a direction histogram, to represent other unique points space constraint of oneself relatively.This algorithm can solve the Point Pattern Matching of non-rigid transformation, most of similar algorithm adopts SPL that conversion is estimated, yet the problem of this class algorithm is the sampling to unique point has dependence, image background can not be too complicated, the test of various algorithms is also mostly based on generated data, use for simple form fit, algorithm performance is relatively good, but the computation complexity of this class algorithm is than higher, each unique point all needs to utilize every other point to calculate in shape hereinafter, so in complex scene, under the situation that the object of identification relatively is difficult to again cut apart, the poor performance of algorithm.

Loose labeling algorithm defines a kind of probability distribution on transformation relation, be optimized with discrete logarithm then, algorithm is actually the process of an iteration, since a thicker coupling, progressively upgrade the matching probability matrix then, this algorithm is easy to converge on local extremum.

Progressively optimized Algorithm also is to utilize global characteristic point to make up a compatible function, and by upgrading compatible function to solve correspondence, this algorithm has identical defective with loose algorithm.Two class algorithms are bigger at feature point set, and algorithm performance is all poor under the many again situation of noise, and can not guarantee that algorithm can restrain in all cases.

The figure matching algorithm is the Point Pattern Matching algorithm of the comparative heat of recent researches.By two point sets to be matched are built into weighted graph, attributed relational graph etc. are searched for compatible part then in two figure, and promptly the part of mating is probably perhaps carried out the subgraph coupling, perhaps shines upon coupling fully, guarantees the majorized function minimum.This class algorithm can be summed up as the graph search problem, is the problem of a NP difficulty, can not guarantee various survey examples are all separated.A lot of algorithms also all are to utilize majorized function to estimate, obtain more excellent separating rather than optimum solution, another defective of figure coupling is to handle two point sets that all contain much noise point, except computing time is long, so this algorithm also is difficult to obtain rational corresponding relation.

List of references related to the present invention has:

[1]A.Goshtasby.Description?and?discrimination?of?planar?shapes?using?shape?matrices.IEEE?Trans.Pattern?Analysis?and?Machine?Intelligence，7：738-743，1985.

[2]S.H.Chang，F.H.Cheng，W.H.Hsu，and?G.Z.Wu.Fast?algorithm?for?point?pattern?matching：Invariant?to?translations，rotations?and?scale?changes.Pattern?Recognition，30：311-320，1997.

[3]S.Belongie，J.Malik，and?J.Puzicha，“Shape?Matching?and?Object?Recognition?Using?ShapeContexts，”IEEE?Trans.Pattern?Analysis?and?Machine?Intelligence，vol.24，no.4，pp.509-522，Apr.2002.

[4]W.J.Christmas，J.Kittler，and?M.Petrou，“Structural?Matching?in?Computer?Vision?UsingProbabilistic?Relaxation，”IEEE?Trans.Pattern?Analysis?and?Machine?Intelligence，vol.17，no.8，pp.749-764，Aug.1994.

[5]R.C.Wilson?and?E.R.Hancock，“Structural?Matching?by?Discrete?Relaxation，”IEEE?Trans.Pattern?Analysis?and?Machine?Intelligence，vol.19，no.6，pp.634-648，June?1997.

[6]S.Gold?and?A.Rangarajan，“A?Graduated?Assignment?Algorithm?for?Graph?Matching，”IEEE?Trans.Pattern?Analysis?and?Machine?Intelligence，vol.18，no.4，pp.377-388，Apr.1996.

[7]Tibe’rio?S.Caetano，Terry?Caelli，Fel?low，IEEE，Dale?Schuurmans，and?Dante?A.C.Barone.Graphical?Models?and?Point?Pattern?Matching.IEEE?Trans.Pattern?Analysis?and?MachineIntelligence，28：1646-1663，2006.

Summary of the invention

The present invention is intended to overcome the deficiencies in the prior art, proposes a kind of method for registering images based on unique point.

The present invention at first proposes a kind of ring type corner sign indicating number model based on unique point k nearest neighbor space structure, the similarity between the sign indicating number of ring type corner is explored and used it for the similarity of the local space of weighing two unique points.Described similar ring type corner sign indicating number has not only been described the similarity of the local space structure between two unique points, can also be used to estimate local mapping.The sign indicating number modelling of ring type corner is on the similarity transformation basis, and the present invention has carried out loose simultaneously to ring type corner sign indicating number simulated condition, makes matching algorithm have robustness to a certain degree affine and view transformation.Similarity between the unique point is determined like the similar length of ring type corner sign indicating number by their associated maximal phases.The present invention at first carries out structure matching according to the similarity of the local space structure of unique point, utilizes the associated part mapping cluster of match point to be optimized coupling then.The result of optimization of matching is used to estimate optimum mapping.Aspect the acceleration matching process, the present invention has avoided a large amount of unnecessary computations by set up index in the search volume, thereby guarantees that matching algorithm has also reached when guaranteeing precision near the requirement of application in real time.

The invention discloses a kind of algorithm of Point Pattern Matching fast and effectively, its process flow diagram is seen accompanying drawing 1.Comprise that the step is following rapid:

Step 1): feature extraction, two images to be matched are carried out feature extraction, and matching algorithm is based on image characteristic point, and coupling extraction in early stage has convergent-divergent, the feature point set of rotation and translation invariance also can have affine and view transformation to a certain degree as input between the point set;

Step 2): similarity is calculated, the ring type corner sign indicating number of the k nearest neighbor structure of extract minutiae and construction feature point, index is set up in the search volume, utilize maximal phase between binary chop and the increment matching algorithm calculated characteristics point like ring type corner sign indicating number, and with maximal phase like the similar length of ring type corner as the similarity between two unique points, determine the unique point that may mate when calculating similarity;

Step 3): structure matching, similarity between the unique point has been described the similarity of the local space structure of unique point in the physical sense, the present invention is according to the similarity size between the unique point, the unique point that may mate is divided into three classes: the unique point class badClass that can not mate, may mate but mate not unique unique point class unknownCkass, the unique point class goodClass that may mate carries out the part mapping to the match point among the goodClass simultaneously and estimates;

Step 4): optimization of matching, utilize the associated part mapping of match point among the g0odClass to carry out cluster, in cluster result, the maximum class of definition element is a maximum kind, the class of element more than second is a time big class.If simultaneously greater than inferior big dvielement number, perhaps maximum kind and the ratio of the element number of inferior big class are greater than certain threshold value greater than certain threshold value for the maximum kind element number, maximum kind will be estimated optimal transformation as matching result;

Step 5): optimal transformation estimates that the present invention supposes being transformed to except similarity transformation between the image, also has affine and view transformation to a certain degree, and the conversion estimation model is an affined transformation.

Particularly,

Two point sets in the step 1 of the present invention are defined as S and T, and size is respectively n ₁, and n ₂, while defining point P _i∈ S, Q _j∈ T.

Ring type corner sign indicating number in the described step 2 is defined as:

As shown in Figure 2, the k nearest neighbor space structure of unique point can pass through a ring type corner representation.For defining ring type corner sign indicating number better, the present invention with the combination at the angle of every limit and its counterclockwise adjacency in the accompanying drawing 2 (a) be defined as a corner sign indicating number (Edge-AngleCode) EAC=(E, θ).Accompanying drawing 2 (a) has K corner sign indicating number (E _i, θ _i), i=0,1 ..., K-1.Also successively they being linked to be a loop configuration then obtains unique point P by counterclockwise sequential encoding to this K corner sign indicating number _cRing type corner sign indicating number, shown in accompanying drawing 2 (b), the present invention defines the length that K is a ring type corner sign indicating number.

The defined ring type of the present invention corner sign indicating number has following characteristic:

1) connectedness: begin to travel through forward from arbitrary corner sign indicating number, all corner sign indicating numbers can be traveled through in order and finish.For ease of expression, the present invention will travel through subscript and be defined as m _t, wherein t represents nonnegative integer, and m _tRepresent that then t gets surplus result to K, do to guarantee that subscript can not crossed the border when the corner sign indicating number is traveled through like this.

2) space constraint: the order of corner sign indicating number has been described the restriction relation that the neighbours of unique point are ordered.Corner sign indicating number order

Neighbours' point has been described in space structure

With neighbours' point

Adjacency in the counterclockwise direction, and with neighbours' point

Neighbours' point at interval in the counterclockwise direction.Therefore the corner sign indicating number is analyzed to be equivalent in proper order the restriction relation of neighbours' point on space structure analyzed.

3) equivalence: same unique point can have K different ring type corner sign indicating number, but they have described the same local space structure of unique point all of equal valuely.By accompanying drawing 2 (a) as can be known, during as first corner sign indicating number, the present invention will obtain different ring type corner sign indicating numbers with different neighbours point, but they are of equal value in the physical sense, have promptly all described identical space structure.The present invention is defined as equivalence to this specific character of ring type corner sign indicating number.

Similar ring type corner sign indicating number definition in the described step 2 is specially:

As shown in Figure 3, if it is proportional that two ring type corner sign indicating numbers begin to satisfy corresponding sides from a certain opposite side, corresponding angles are proportional and equal 1, the present invention just says that these two ring type corner sign indicating numbers are similar, the length of one of them ring type corner sign indicating number is defined as the similar length of similar ring type corner sign indicating number, and the space structure of similar ring type corner sign indicating number correspondence is defined as similar space structure.Similar ring type corner sign indicating number shown in the accompanying drawing 3 is from (E ₀, E ' ₀) begin to satisfy following formula:

$\frac{\left|{E^{\&prime;}}_0\right|}{\left|{\mathrm{<figure-callout\; id="0"\; label="E"\; filenames="A20091005245300133.png,A20091005245300141.png"\; state="\{\{state\}\}">E</figure-callout>}}_0\right|}=\frac{\left|{E^{\&prime;}}_1\right|}{\left|{\mathrm{<figure-callout\; id="1"\; label="E"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">E</figure-callout>}}_1\right|}=K=\frac{\left|{E^{\&prime;}}_{K-1}\right|}{\left|E_{K-1}\right|}$

$\frac{{{\mathrm{\&theta;}}^{\&prime;}}_0}{{\mathrm{\&theta;}}_0}=\frac{{{\mathrm{\&theta;}}^{\&prime;}}_1}{{\mathrm{\&theta;}}_1}=K=\frac{{{\mathrm{\&theta;}}^{\&prime;}}_{K-1}}{{\mathrm{\&theta;}}_{K-1}}=1$

Be calculation procedure two described similar ring type corner sign indicating numbers, the present invention has also defined the add operation of two adjacent corner sign indicating numbers, is specially:

$(E_{m_i},{\mathrm{\&theta;}}_{m_i})+\left(E_{m_{(i+1)}},{\mathrm{\&theta;}}_{m_{(i+1)}}\right)=(E_{m_i},{\mathrm{\&theta;}}_{m_i}+{\mathrm{\&theta;}}_{m_{(i+1)}})$

The result of two adjacent corner sign indicating number add operations still is a corner sign indicating number, and the limit of new corner sign indicating number is the limit of first corner sign indicating number, and the angle of new corner sign indicating number then is the angle sum of two addition corner sign indicating numbers.In fact the physical significance of phase add operation has been represented two adjacent neighbours' points of counter clockwise direction are merged, and is about to second neighbours' point deletion, then the angle of the counterclockwise adjacency in article one limit is upgraded.

Be calculation procedure two described similar ring type corner sign indicating numbers, the present invention has also defined the compare operation of two corner sign indicating numbers, if two corner sign indicating number (E ₁, θ ₁), (E ' ₁, θ ' ₁) satisfy

$\mathrm{abs}(\frac{\left|{E^{\&prime;}}_1\right|/\left|{\mathrm{<figure-callout\; id="1"\; label="E"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">E</figure-callout>}}_1\right|}{\left|{E^{\&prime;}}_2\right|/\left|{\mathrm{<figure-callout\; id="2"\; label="E"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">E</figure-callout>}}_2\right|}-1.0)<\mathrm{\&epsiv;}$ $\mathrm{abs}(\frac{{{\mathrm{\&theta;}}^{\&prime;}}_1}{{\mathrm{\&theta;}}_1}-1.0)<\mathrm{\&epsiv;}$

Wherein ε is the error control factor, approaches 0 positive number, (E ₂, θ ₂), (E ' ₂, θ ' ₂) be respectively (E ₁, θ ₁), (E ' ₁, θ ' ₁) the corner sign indicating number of next adjacency, the present invention is regulation (E just ₁, θ ₁), (E ' ₁, θ ' ₁) compare operation do not need add operation, the physical significance that does not need add operation be the expression by E ₁, θ ₁, E ₂The triangle that constitutes with by E ' ₁, θ ' ₁, E ' ₂The triangle that constitutes is similar.

Described step 2 similarity is calculated and is specially:

(1) set up the search volume, the ring type corner sign indicating number of the k nearest neighbor of extract minutiae and construction feature point, the search volume is that impact point is concentrated the n on the ring type corner sign indicating number of all unique points ₂K corner sign indicating number set up index with acceleration search to these corner sign indicating numbers by angle.Synoptic diagram is seen in the middle of the accompanying drawing 4 and the right two tabulation lattice.

(2) determine object search, the coupling purpose is to find match point.Therefore the single object search is P _iEach associated corner sign indicating number of ring type corner sign indicating number, with regard to P _i, total K object search (E _k, θ _k), shown in accompanying drawing 4 left side forms.

(3) determine search strategy, the present invention is at first to (E _k, θ _k) utilize the binary search corner sign indicating number (E that angle is identical in the seek error allowed band in the search volume _n, θ _n), establishing characteristic of correspondence point is Q _jIf find, the present invention also can be at other identical (E of linear search angle in the error allowed band before and after the sign indicating number of finding corner _n, θ _n).If (E _k, θ _k) and (E _n, θ _n) comparative result do not need add operation, then respectively with (E _k, θ _k) and (E _n, θ _n) be the ring type corner sign indicating number L of two equivalences of dot generation _P={ A ₀, A ₁..., A _(K-1)And L _Q={ B ₀, B ₁..., B _(K-1), A wherein _k, B _nExpression corner sign indicating number, note E _k ^A, θ _k ^A, E _n ^B, θ _n ^BRepresent A respectively _k, B _nBian Hejiao.Press their maximal phase of increment matching algorithm calculating then like ring type corner sign indicating number.The increment matching algorithm is as follows: at first to A ₀, B ₀Compare operation, if θ ₀ ^A, θ ₀ ^BUnequal, the next corner sign indicating number that the corner sign indicating number that then angle is little is adjacent carries out add operation and generates new corner sign indicating number; If θ ₀ ^A, θ ₀ ^BEquate, then verify θ ₀ ^A, θ ₀ ^BAdjacent corresponding sides (E ₀ ^A, E ₀ ^B) and (E ₁ ^A, E ₁ ^B) whether proportional, if proportional, then with A ₀, B ₀Be retained in respectively in the similar ring type corner sign indicating number of having tried to achieve (just having begun to be sky), if disproportionate, then with this A ₀, B ₀Carry out the new corner sign indicating number of add operation generation with adjacent separately next corner sign indicating number, next compare operation is carried out in circulation then, and cyclic process is till one of them ring type corner sign indicating number has traveled through.False code is expressed as follows:

Suppose t ₁And t ₂Be two loop variables, t ₁=0, t ₂=0.Make temp _A=A _k, temp _B=B _n, rate=E _n ^B/ E _k ^A, then the increment matching algorithm is as follows:

while(t ₁＜K?and?t ₂＜K){

$\mathrm{if}({\mathrm{\&theta;}}_{{\mathrm{temp}}_A}<{\mathrm{\&theta;}}_{{\mathrm{temp}}_B})\{{\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">t</figure-callout>}}_1={\mathrm{<figure-callout\; id="1"\; label="t"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">t</figure-callout>}}_1+1,{\mathrm{temp}}_A={\mathrm{temp}}_A+A_{m(i+t1)}.\}$

$\mathrm{else\; if}({\mathrm{\&theta;}}_{{\mathrm{temp}}_A}>{\mathrm{\&theta;}}_{{\mathrm{temp}}_B})\{{\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">t</figure-callout>}}_2={\mathrm{<figure-callout\; id="2"\; label="t"\; filenames="A20091005245300132.png,A20091005245300133.png"\; state="\{\{state\}\}">t</figure-callout>}}_2+1,{\mathrm{temp}}_B={\mathrm{temp}}_B+B_{m(i+t2).}\}$

else{

t ₁＝t ₁+1，t ₂＝t ₂+1.

$\mathrm{if}(\mathrm{rate}=={E^B}_{m(n+t_2)}/{E^A}_{m(k+t_1)})$

{Add?temp _AintoL _k，add?temp _B?into?L _n，temp _A＝A _m(k+t1)，temp _B＝B _m(n+t2).}

else{temp _A＝temp _A+A _m(k+t1)，temp _B＝temp _B+Bm _(n+t2).}}}

The similar ring type corner sign indicating number that only keeps similar length maximum in the search procedure when the ring type corner of similar length maximum sign indicating number has when a plurality of, all keeps.With maximal phase like the similar length of ring type corner sign indicating number as the similarity between two unique points, notice that the similarity of all two unique points not calculated in the search procedure is 0.

(4) the definite unique point that may mate, the present invention defines S (A _k, B _n)={ L _k, L _nRepresent with E _k ^A, E _n ^BBe the similar ring type corner sign indicating number of initial corresponding sides, L _k, L _nRepresent similar ring type corner sign indicating number.Make S (A _k, B _n) similar length be len ^Ij(k, n), unique point P then _i, Q _jSimilarity be:

Similarity(P _i，Q _j)＝arg?max?len ^ij(k，n)?k，n∈[0，K-1]

If Q _j ^*Satisfy

${Q_j}^{*}=\arg {\max }_{Q_j\&Element;T}\mathrm{similarity}(P_i,Q_j)$

The present invention just thinks Q _j ^*Be P _iPossible match point, P in the computation process _iHave a plurality of Q that may mate _j ^*, all keep.

Step 3 structure matching of the present invention is specially:

If P _iThe Q that may mate _j ^*There is NO, then to possibility match point (P _i, Q _* ^j) be classified as follows:

$(P_i,{Q^{*}}_j)\&Element;\left\{\begin{array}{ccc}\mathrm{badClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)<3\\ \mathrm{unknownClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)\&GreaterEqual;3\mathrm{andNO}>1\\ \mathrm{goodClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)\&GreaterEqual;3\mathrm{andNO}=1\end{array}\right.$

Class condition is explained as follows: (the P among the badClass _i, Q _* ^j) similarity all less than 3, when K is bigger,, (P is described as K=15 _i, Q ^* _j) the local space textural difference very big, P _iOr Q ^* _jMay be noise spot, the possibility of coupling be very little; (P among the unknownClass _i, Q ^* _j) similarity all more than or equal to 3, when K is bigger,, (P is described as K=15 _i, Q ^* _j) the local space texture ratio more similar, but P _iThe Q that may mate ^* _jMore than one, this explanation only can't be determined P by the similarity of local space structure _iWith which Q ^* _jCoupling needs subsequent authentication to determine; (P among the goodClass _i, Q ^* _j) similarity all more than or equal to 3, when K is bigger,, (P is described as K=15 _i, Q ^* _j) the local space texture ratio more similar, and P _iThe Q that may mate ^* _jHas only one, (P _i, Q ^* _j) correct possibility of mating is very big, and similarity is big more, illustrates that the local space structure is similar more, the possibility of coupling is big more.

According to above-mentioned analysis, the present invention is with (the P among the goodClass _i, Q ^* _j) as the result of structure matching, the present invention simultaneously is also to the (P among the goodClass _i, Q ^* _j) carry out the local similar conversion and estimate,, note in local similar conversion estimation procedure (P _i, Q ^* _j) the one group respective neighbours point related with them make up and then estimate one group of local similar conversion, if (P _i, Q ^* _j) similarity be ls, then can estimate the conversion of ls group local similar, matching algorithm is got the average of ls group local similar conversion as the final local similar conversion of estimating.

Described step 4, optimization of matching is specially:

(1) to (P among the goodClass _i, Q ^* _j) sort by similarity is descending;

(2) to arranging (the P of preface _i, Q ^* _j) scan, if (P _i, Q ^* _j) not by the cluster mistake, then with (P _i, Q ^* _j) be the class center C _i, (P _i, Q ^* _j) associated part mapping Ti is as sorter, do not verified among goodClass and the unknownClass by the possible match point of cluster and related respective neighbours point.Cluster process is as follows: at first with (P _i, Q ^* _j) related respective neighbours point is included into such, because sorter T _iCome out by these respective neighbours point estimation, utilize sorter to verify among other goodClass and the unknownClass the not possible match point of cluster and related respective neighbours point then, if the point of being verified is to satisfying sorter, then it is carried out mark, no longer carry out subsequent treatment and it is included into such, otherwise it is ignored.If a pair of point that wherein needs to verify is for (P Q), establishes P through T _iPoint after the mapping is T _i(P), if T _i(P) can accept in error range with the Euclidean distance of Q, the present invention thinks that then (P Q) meets T _i, with (P Q) is classified as C _i

Be formulated as:

|T _i(P)，Q?|＜ε _T

ε _TBe the error control factor.Attention is every couple of (P in cluster process _i, Q ^* _j) only by cluster once, i.e. (P _i, Q ^* _j) belong to and can only belong to a class.Cluster process must to all not cluster cross other may match points and related respective neighbours point verify.The beginning condition of cluster is also to exist among the goodClass not have cluster to cross (P next time _i, Q ^* _j).

(3) cluster result is analyzed,, it must be deleted because generally there is erroneous matching among the structure matching result.In cluster result, the maximum class of definition element is a maximum kind, and the class of element more than second is a time big class, a pair of match point of each element representation in the class.Generally, the associated part mapping of the unique point of correct coupling also is correct, and all correct match points all can gather in the error allowed band into a class, and the point of respective class is very many to meeting; Otherwise the associated part of the point of erroneous matching mapping randomness is very big, and element is considerably less in the corresponding class.Therefore to be optimized coupling as follows in the present invention: if the ratio of maximum kind and the element number of inferior big class is greater than certain threshold value, perhaps the maximum kind element number is greater than inferior big class, the maximum kind element number is greater than certain threshold value simultaneously, the present invention will export maximum kind as corresponding relation, carrying out optimal transformation estimates, otherwise the present invention thinks that then two point set correlativitys of coupling are too poor, can't mate.

Description of drawings

Fig. 1; Process flow diagram of the present invention.

Fig. 2: unique point P _cKNN space structure figure and corresponding ring type corner sign indicating number.

Fig. 3; Similar ring type corner sign indicating number and corresponding similar space structure.

Fig. 4: similarity is calculated synoptic diagram.

Fig. 5: two images to be matched.

Fig. 6: the matching result of two width of cloth images shown in Figure 5.

Embodiment

Below provide instantiation to further specify application of the present invention.

Embodiment 1

Match objects is seen two width of cloth images shown in the accompanying drawing 5.Two width of cloth image takings are in different time, and different angles have only partial content identical, and the image size is 440 * 330.

Moving equipment of the present invention is Gateway T6307c notebook, Intel 2core 1.6G, 1G internal memory.The present invention has platform transplantation, and the specific implementation platform is GCC environment under the Centos and the Visual C++2005 environment under the Windows XP/Server 2003.Operational efficiency GCC is better than Visual C++2005.

The present invention stipulates the Euclidean distance error ε between equal error ε=0.1, two pixel of two scalars _T=10, define two kinds of errors and be intended to calculate similar ring type corner sign indicating number and cluster.Neighbours' number of the unique point of the present invention simultaneously is set to K=15.

Concrete performance is as follows:

Step 1: obtain two feature point sets as input from two images to be matched, unique point only need comprise that two-dimensional coordinate information gets final product.The present invention has adopted single scale Harris Corner Detection Algorithm, Harris response coefficient α=0.06, difference yardstick σ D=3, level and smooth yardstick σ _I=7, it is 3 * 3 that local non-maximum value suppresses window, and inhibiting value is 0.01 times of overall maximum angle point response.Two width of cloth images extract 189 and 208 unique points respectively.Corresponding point set is designated as S and T, n respectively ₁=189 and n ₂=208.

Step 2: similarity is calculated, and 15 neighbours that extract each unique point are according to the ring type corner sign indicating number of the definition calculated characteristics point of ring type corner sign indicating number.In computation process, avoid removing Z-operation.When scalar is done division, as required, add or deduct a very little positive number, in order to avoid program crashing.

According to the definition of similar ring type corner sign indicating number, calculate the similarity of two unique points to be matched, according to classifying rules, determine the element among goodClass and the unknownClass, calculate the associated part mapping of element among the goodClass simultaneously.Concrete steps are as follows:

(1) sets up the search volume.208 * 15 corner sign indicating number (E to the ring type corner sign indicating number of 208 unique points among the T _n, θ _n) sort to set up index according to angular dimension.

(2) determine object search.For obtaining P _iThe point Q that may mate ^* _j, need be to P _iRing type corner sign indicating number on 15 corner sign indicating number (E _k, θ _k) (k=0,1 ..., 15) search for respectively.

(3) determine search strategy.Select each (E successively _k, θ _k), search for by angle in the search volume, if search for successfully, the corner sign indicating number (E that is searching then _n, θ _n) front and back continuation linear search other corners sign indicating number (E that angle equates in error range _n, θ _n), if (E _k, θ _k) and (E _n, θ _n) comparative result do not need add operation, then respectively with (E _k, θ _k) and (E _n, θ _n) be the ring type corner sign indicating number of two equivalences of dot generation and utilize maximal phase that the increment matching algorithm calculates them like ring type corner sign indicating number.

(4) the definite unique point that may mate, P _i15 (E _k, θ _k) after all search finishes, keep the similar ring type corner sign indicating number of similar length maximum in the computation process, with the maximal phase that keeps another unique point Q like the sign indicating number association of ring type corner _jAs P _iThe point Q that may mate ^* _j, then all preserve if any a plurality of.

Step 3: structure matching, 189 unique points among the S are classified according to similarity and the unique point number that may mate.Keep goodClass and unknownClass.And according to the mapping of the part of elements correlation among the similarity transformation Model Calculation goodClass.The similarity transformation model is as follows:

X＝sxcosθ-sysinθ+t _x

Y＝sxsinθ+sycosθ+t _y

S wherein, θ is respectively the zoom factor and the anglec of rotation, t _x, t _yIt is respectively the translational movement of x and y direction.(x is y) with (X Y) is the unique point coordinate of conversion front and back respectively.Note in the local similar transformation calculations process, just the possible match point one group respective neighbours point related with them can be estimated one group of conversion coefficient, if so P _i, Q ^* _jSimilarity is ls, then can estimate the conversion of ls group local similar, and the present invention organizes ls the average of local similar conversion as the final local similar conversion of estimating.

Step 4: optimization of matching, sort by similar length from big to small to the element among the goodClass in the step 3.Because similar length is big more, the local space structure of two unique points of expression is similar more, and the possibility of coupling is also big more, from the P of similar length maximum _i, Q ^* _jThe beginning cluster just can guarantee that also the cluster effect is good more, and the matching effect that obtains is also good more.Therefore the cluster a pair of unique point of at first selecting not gather the similarity maximum of class begins.Carry out cluster by clustering rule then.Splicing at the large scale scene image is used with identification, establishes s ₁, s ₂Be the element number of maximum kind and time big class, if $s_1\&GreaterEqual;2s_2$ Perhaps s ₁＞s ₂, s ₁＞23, algorithm will be used for optimal transformation and estimate the output of the element in the maximum kind as the optimization of matching result.Otherwise algorithm will think that coupling is unsuccessful.At the example of accompanying drawing 5, cluster result one has 88 classes, and wherein the maximum kind number of elements is 33, and inferior big dvielement number is 6.By observing, the maximum kind element all is correct the coupling in error range, and inferior big dvielement all is erroneous matching in error range.In accompanying drawing 5 example illustrated, maximum kind is 5.5 with inferior big dvielement number ratio, satisfies condition, and the match is successful.The element of maximum kind will carry out optimal transformation and estimate as corresponding relation.

Step 5: conversion estimates that because generally, the conversion between the scene image often has affine or view transformation to a certain degree, so the present invention adopts the conversion between the affined transformation model estimation scene image.The affined transformation model is as follows:

X＝ax+by+t _x

Y＝cx+dyt _y

A wherein, b, c, d are respectively rotations, convergent-divergent and stretching factor, t _x, t _yIt is respectively the translational movement of x and y direction.(x is y) with (X Y) is the unique point coordinate of conversion front and back respectively.To all match points,, estimate to obtain optimal transformation thereby carry out least square fitting by the linear system of equations of affined transformation model group.At the example of accompanying drawing 5, optimal transformation is $\left[\begin{array}{ccc}1.04237& 0.00366& -333.03\\ 0.01088& 1.02362& -10.1731\end{array}\right],$ Optimal transformation is applied to matching image, obtains matching result, see accompanying drawing 6.The result shows that the present invention has only the identical image of partial content to be stitched together well two width of cloth in error range.

Claims (9)

1, a kind of method for registering images based on unique point is characterized in that, may further comprise the steps:

Step 1): feature extraction,

Two imagery exploitation feature point extraction operators subject to registration are carried out feature point extraction, and two feature point sets that obtain are defined as the input data of method;

Step 2): similarity is calculated,

The ring type corner sign indicating number of the k nearest neighbor structure of extract minutiae and construction feature point, index is set up in the search volume, utilize maximal phase between binary chop and the increment matching algorithm calculated characteristics point like ring type corner sign indicating number, maximal phase is defined as two similarities between the unique point like the similar length of ring type corner, determines when calculating similarity that the unique point that may mate is right;

Step 3): structure matching,

Count with mating according to the similarity size, unique point is shone upon estimation to classifying and carrying out the part;

Step 4): optimization of matching,

Utilize the structure matching result to carry out cluster, determine the unique point corresponding relation;

Step 5): the optimal transformation estimation,

Utilize the affined transformation model that least-squares estimation is carried out in the mapping between two width of cloth images.

2, method according to claim 1, it is characterized in that, described step 2) it is defined as the ring type corner sign indicating number in: in the K bar limit of the k nearest neighbor space structure of constitutive characteristic point and K angle being made of adjacent edge, the combination at the angle of every limit and it is counterclockwise or clockwise adjacency is defined as the corner sign indicating number, K corner sign indicating number is pressed the ring type corner sign indicating number of counterclockwise or clockwise sequential encoding constitutive characteristic point.

3, method according to claim 1 and 2, it is characterized in that, described ring type corner sign indicating number is under the jurisdiction of certain unique point, in the K that is comprised the corner sign indicating number, any two corner sign indicating numbers all are communicated with, each corner sign indicating number also is under the jurisdiction of certain unique point, corresponding K the ring type corner sign indicating number that starting point is different of each unique point, and K is defined as the length of ring type corner sign indicating number.

4, method according to claim 1 and 2, it is characterized in that, described step 2) the similar ring type corner sign indicating number in is defined as two, and to satisfy corresponding sides proportional, corresponding angles are proportional and equal 1 ring type corner sign indicating number, the length of single ring type corner sign indicating number is defined as similar length, and similar length is defined as the similarity between two unique points that similar ring type corner sign indicating number is subordinate to.

5, method according to claim 1 and 2 is characterized in that, the add operation of two adjacent corner sign indicating numbers is defined as:

$(E_{m_i},{\mathrm{\&theta;}}_{m_i})+(E_{m_{(i+1)}},{\mathrm{\&theta;}}_{m_{(i+1)}})=(E_{m_i},{\mathrm{\&theta;}}_{m_i}+{\mathrm{\&theta;}}_{m_{(i+1)}})$

m _tExpression nonnegative integer t gets surplus result to K, With

Represent two adjacent corner sign indicating numbers.

6, method according to claim 1 and 2 is characterized in that, the compare operation of two corner sign indicating numbers is that genuine conditional definition is:

$\mathrm{abs}(\frac{\left|{E^{\&prime;}}_1\right|/\left|E_1\right|}{\left|{E^{\&prime;}}_2\right|/\left|E_2\right|}-1.0)<\mathrm{\&epsiv;}$ $\mathrm{abs}(\frac{{{\mathrm{\&theta;}}^{\&prime;}}_1}{{\mathrm{\&theta;}}_1}-1.0)<\mathrm{\&epsiv;}$

Wherein ε is the error control factor, approaches 0 positive number, (E ₂, θ ₂), (E ' ₂, θ ' ₂) be respectively corner sign indicating number (E ₁, θ ₁), (E ' ₁, θ ' ₁) the corner sign indicating number of next adjacency.

7, according to claim 1 or 2 or 4 described methods, it is characterized in that described step 2) in similarity calculate by following step:

(1) set up the search volume, define two point sets and be defined as S and T, size is defined as n respectively ₁And n ₂, define P simultaneously _i∈ S, Q _j∈ T, the k nearest neighbor of extract minutiae is also pressed the ring type corner sign indicating number that defines construction feature point, and the search volume is defined as the pairing n of point set T ₂N on the sign indicating number of individual ring type corner ₂K corner sign indicating number is to n ₂K corner sign indicating number set up index by angular dimension;

(2) determine object search, the single object search is defined as P _iRing type corner sign indicating number on each corner sign indicating number, object search is K;

(3) determine search strategy,

Definition (E _k, θ _k) be P _iA corner sign indicating number, utilize binary search and linear search all identical corner sign indicating number (E of angle in the seek error allowed band in the search volume _n, θ _n), definition (E _n, θ _n) unique point that is subordinate to is Q _j

(4) utilize the increment matching algorithm to calculate similar ring type corner sign indicating number,

With (E _k, θ _k) and each (E _n, θ _n) compare,

As (E _k, θ _k) and (E _n, θ _n) compare operation be true time, respectively with (E _k, θ _k) and (E _n, θ _n) be two ring type corners of dot generation sign indicating number L _P={ A ₀, A ₁..., A _(K-1)And L _Q={ B ₀, B ₁..., B _(K-1), A wherein _k, B _nExpression corner sign indicating number, note E _k ^A, θ _k ^A, E _n ^B, θ _n ^BRepresent A respectively _k, B _nBian Hejiao,

At first compare A ₀, B ₀,

Work as θ ₀ ^A, θ ₀ ^BWhen unequal, the next corner sign indicating number that the corner sign indicating number that angle is little is adjacent carries out add operation,

Work as θ ₀ ^A, θ ₀ ^BWhen equating, checking θ ₀ ^A, θ ₀ ^BAdjacent corresponding sides (E ₀ ^A, E ₀ ^B) and (E ₁ ^A, E ₁ ^B) whether proportional, if proportional, then with A ₀, B ₀Be retained in respectively in the similar ring type corner sign indicating number of having tried to achieve, then with A ₀, B ₀Carry out add operation with adjacent separately next corner sign indicating number,

Carry out the sign indicating number compare operation of next corner by the same rule circulation, till one of them ring type corner sign indicating number has traveled through;

(4) determine the possibility match point, definition S (A _k, B _n)={ L _k, L _nRepresent with E _k ^A, E _n ^BBe the similar ring type corner sign indicating number of initial corresponding sides, L _k, L _nRepresent similar ring type corner sign indicating number, definition S (A _k, B _n) similar length be len ^Ij(k, n), unique point P then _i, Q _jSimilarity be:

Similarity(P _i，Q _j)＝arg?max?len ^ij(k，n) k，n∈[0，K-1]

Work as Q _j ^*Satisfy

${Q_j}^{*}={\arg \max }_{Q_j\&Element;T}\mathrm{similarity}(P_i,Q_j)$

The time, definition Q _j ^*Be P _iPossible match point, Q _j ^*Number be defined as NO.

8, according to claim 1 or 2 or 4 or 7 described methods, it is characterized in that, the structure matching in the described step 3), unique point is defined as class condition:

$(P_i,{Q^{*}}_j)\&Element;\left\{\begin{array}{ccc}\mathrm{badClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)<3\\ \mathrm{unknownClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)\&GreaterEqual;3\mathrm{andNO}>1\\ \mathrm{goodClass}& \mathrm{if}& \mathrm{similarity}(P_i,{Q^{*}}_j)\&GreaterEqual;3\mathrm{andNO}=1\end{array}\right.$

(P among the goodClass _i, Q ^* _j) be defined as the structure matching result, to (the P among the goodClass _i, Q ^* _j) adopt the similarity transformation model to carry out part mapping estimation.Definition (P _i, Q ^* _j) similarity be ls, (P _i, Q ^* _j) put combination with the every couple of neighbour and estimate the conversion of ls group local similar altogether, the average of definition ls group local similar conversion is as the final part mapping of estimating.

9, according to claim 1 or 2 or 4 or 8 described methods, it is characterized in that the optimization of matching of described step 4) is by following step;

(1) to (P among the goodClass _i, Q ^* _j) sort by similarity is descending;

(2) to arranging (the P of preface _i, Q ^* _j) carry out descending scanning by the similarity size, as (P _i, Q ^* _j) not by the cluster mistake, then with (P _i, Q ^* _j) be the class center C _i, (P _i, Q ^* _j) associated part mapping T _iAs sorter, among goodClass and the unknownClass by the unique point of cluster to and related respective neighbours point do not verify;

At first with (P _i, Q ^* _j) related respective neighbours point directly is included into such, utilize sorter to verify among other goodClass and the unknownClass the not possible match point of cluster and related respective neighbours point then, satisfy sorter T in the error range when the point of being verified closes to tie up to volume coordinate _iThe time, then it is carried out mark, it is included into such, do not carry out subsequent treatment, otherwise it is ignored;

(3) analyze cluster result,

In the cluster result, the class definition that element is maximum is a maximum kind, and the class of element more than second is a time big class; The maximum kind element number is defined as S ₁, inferior big dvielement number is defined as S ₂, when one of following condition satisfied, maximum kind was estimated optimal transformation as the unique point corresponding relation,

1. S ₁＞t ₁S ₂, 2. S ₁＞S ₂While S ₁＞t ₂,

T wherein ₁, t ₂Be certain threshold level, by the kind decision of image to be matched greater than 1.

Images