CN104182966A - Automatic splicing method of regular shredded paper - Google Patents

Abstract

The invention belongs to an image processing technology and particularly relates to an automatic splicing method of regular shredded paper. The technical scheme of the invention is achieved through six steps of (1) preparing an image data set and preprocessing; (2) classifying the shredded paper according to Chinese and English as well as a single face and two faces; (3) extracting local regional characteristics of each image, such as a position and a grey level of a boundary pixel point of the shredded paper as well as an upper (lower) boundary height; enlarging an extraction range of the characteristics of the English shredded paper, wherein the supplementary features include a row height of the English shredded paper, a horizontal position of the English shredded paper and a vertical spacing of the English shredded paper; (4) reclassifying the shredded paper according to feature values extracted in the step (3); (5) carrying out local matching, line matching and column matching on the shredded paper; and (6) restoring the matched images. According to the automatic splicing method provided by the invention, a large amount of the shredded paper can be more accurately spliced.

Description

A kind of regular shredded paper method for automatically split-jointing

Technical field

The invention belongs to the application of image processing techniques, be specifically related to a kind of regular shredded paper method for automatically split-jointing.

Background technology

Shredded paper splicing is the Yi Ge important research branch of digital image processing techniques, and it is to exist each other the shredded paper of lap to carry out space coupling aligning by one group, thereby carries out the seamless spliced image complete, wide visual angle scene that obtains.

Shredded paper automatic Mosaic recovery technique restores at judicial material evidence, historical document reparation and military information are obtained etc., and there is important application in field.In recent years, along with the announcement that this tower western language part of Germany recovers engineering, the research of shredded paper file reset technology has caused widely to be paid close attention to.

The key that shredded paper splicing must complete is the matching technique of fragment.The splicing of the broken file of tradition is more to use the edge shape of fragment extract its contour curve and utilize computerized algorithm to splice.Nowadays along with the widespread use of shredder, in increasing broken scraps of paper Bonding Problem, the edge shape of shredded paper is all roughly the same, and edge shape splicing is no longer applicable.For the shredded paper of regular shape, be the word content comprising according to sheet edge, by image registration computing, determine the parameter on shredded paper border, fragment is mated, finally realize seamless spliced.But in the middle of practical application, scraps of paper quantity to be spliced is larger, the sheet edge quantity with similar Word message is also just larger, and similarity degree is higher.And the resolution of computer scanning formation digital picture has certain limitation, therefore, in splicing, there will be a certain amount of wrong splicing.The effect that desirable splicing will reach is " zero error ".With regard to the existing state of the art, existing shredded paper joining method is mostly directed to irregular shape, and the method that can effectively be applied to the splicing of the large-scale wide cut rule scraps of paper is comparatively rare.

The key problem in technology that improves shredded paper automatic Mosaic quality is to obtain how in high quality word or the image information on shredded paper.In general, the quantity of information on fragment is less, and splicing probability wrong or even that cannot splice is larger.Therefore up to now, in this technical field, shredded paper image being carried out to automatic Mosaic process wishes to access final high-quality wide cut shredded paper splicing and has the technical difficulty that paper is larger.

Summary of the invention

A kind of method that the object of this invention is to provide regular shredded paper automatic Mosaic, can splice a large amount of a scrap of papers more exactly.

The present invention is achieved by the following technical solutions, mainly comprises following six steps:

1. the preparation of image data set and pretreated concrete steps comprise:

1.1 by a scrap of paper from left to right, number consecutively from top to bottom, be designated as 1,2,3n; If desired distinguish pros and cons, front is designated as a1, a2, a3an; Reverse side is designated as b1, b2, b3bn;

1.2 by image digitazation, usings pixel as least unit, and extracts gray-scale value and the position of each pixel, sets up Jacobian matrix;

1.3 carry out value by image: the point that gray-scale value is " 0 " is black color dots, and the point that gray-scale value is " 255 " is white point, and between " 0 " and " 255 " is Grey Point;

1.4 denoising points: because raw information is all continuous simulating signal, digitized processing image later should be also a discontinuous point image with continuous trend.For same color dot, surround the situation of heterochromatic point completely, by the same color around that changes into of the color of heterochromatic point;

2. pair shredded paper integral body is classified, by Chinese and English, single two-sided 4 kinds of situations that are divided into: Chinese one side, two-sided, the English one side of Chinese, English two-sided;

3. extract respectively the feature of every width image local area, these features comprise: the position of a scrap of paper boundary pixel point and gray-scale value, upper (lower) boundary height; Extraction scope to English a scrap of paper feature expands, and supplementary features comprise: the row of English a scrap of paper is high, the horizontal level of English a scrap of paper is, the line space of English a scrap of paper;

The method of feature extraction is specific as follows:

I) position of the outermost pixel of a scrap of paper and gray-scale value:

Definition a scrap of paper the most left (right side) end one row pixel is left (right side) border, and pushing up (end) end one-row pixels point is most upper (lower) border, extracts position and the gray-scale value of each boundary pixel point;

Ii) upper (lower) boundary height:

According to the whether complete white of the up-and-down boundary of each fragment, be divided into white border height and the large class of black border height two.Concrete sorting technique is as follows:

The lowermost end of fragment of take is x axle, take the fragment left side perpendicular to x axially upper be y axle, the joining of x axle and y axle is that initial point is set up coordinate system, and each pixel on picture is done to projection to y axle.As shown in Figure 1.The projection of black or Grey Point is designated as once effectively projection, and projection number of times adds 1, and the projection of white point is invalid, and projection number of times does not change.Distance between record and initial point is the projection number of times f (h) on the subpoint of h pixel.

$g\left(h\right)=\left\{\begin{array}{c}0,f\left(h\right)<n/10\\ 255,f\left(h\right)\&GreaterEqual;n/10\end{array}\right.$

When projection number of times f (h) be less than the total pixel n of this row 1/10 time, the gray-scale value g (h) that puts h on y axle is designated as to " 0 "; When projection number of times f (h) be more than or equal to the total pixel n of this row 1/10 time, the gray-scale value g (h) of a h is designated as to " 255 ".

On axis of projection, from the coboundary of fragment, add up downwards successively, until there is the point that color is different.This section is highly coboundary height, and lower boundary height is as the same.

Iii) row of English a scrap of paper is high:

The height of English alphabet and in same a line shared position height roughly the same, therefore, according to step I) mode carry out projection, the interval that gray-scale value is " 1 " is alphabetical valid interval, the height of definition valid interval for row high;

Iv) horizontal level of English a scrap of paper:

Through step I) after projection, the up-and-down boundary between effective projection in zone of letter, is called the place horizontal level of this row letter, in order to determine this row letter position on a scrap of paper apart from the distance at a scrap of paper top;

V) line space:

Extract vertical range between two horizontal levels as line space;

4. the feature set of extracting according to step 3, fragment is classified:

Concrete steps are as follows:

I) according to sheet edge, whether there is strokes of characters information, shredded paper is divided three classes: up-and-down boundary fragment, border, left and right fragment and middle fragment;

Ii) according to line space feature, respectively above-mentioned three class a scrap of papers further to be classified, identical line space is divided into a class;

Iii) according to upper (lower) boundary height, to step I) formed three class fragment collection classify, and the fragment that upper (lower) boundary height is identical or close is divided into same fragment collection:

Divide classification and need to follow certain condition:

(1) number of tiles of each class must be equal to or slightly less than the rip cutting number of times of paper;

(2) with other classifications highly separately, if quantity is less than 1/5 of number of tiles of all categories, be not independently a classification;

(3) continuous several classifications are classified as same class highly mutually;

(4) final classification adds up to the crosscut number of times of paper;

(5) if still cannot determine classification, with same method, bottom level is carried out to auxiliary judgment again.

Iv) utilizing horizontal level, to step I i) formed each fragment collection further classifies, and the fragment in same level position is divided into a class;

5. the concrete steps that pair fragment mates:

5.1 pairs of fragments carry out local matching, are the couplings between two fragments, take left and right coupling below as example:

I. define X _ijbe the gray-scale value that i opens the capable pixel of j on fragment right margin, definition Y _{i ' j}it is the gray-scale value (i ≠ i ') of the capable pixel of j on i ' fragment left margin.Judge that the key of whether mating is X _ijand Y _{i ' j}between matching degree, the feature set that step 3 is extracted, is characterized as benchmark with right margin, definition criterion is:

X _ijfor white, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}occur greyish white black three looks and entirely for black for normal, can mate;

X _ijfor grey, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}occur that random colour is normally, can mate;

X _ijfor black, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}be not that white is normal entirely, can mate;

All the other situations are undesired, can not mate.

X _ijwith Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}relation as shown in Figure 2.

Wherein: X _ij: i opens the gray-scale value of the capable pixel of j of Far Left one row of paper slip;

Y _{i ' j}: the gray-scale value of the capable pixel of j of rightmost one row of i ' paper slip;

Edge following algorithm idiographic flow is as follows:

(1) choose fragment i and i '

(2) suppose that fragment i and i ' mate mutually;

(3) read the capable pixel X of fragment i right margin j _ijgray-scale value;

(4) the capable pixel Y of j-1, j, j+1 of scanning fragment i ' left margin _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}, judge whether it is white entirely;

(5) if be white entirely, and exceed line range, after j=j+1, return (3);

(6) if be not white entirely, j=j+1, reads next line, judgement X _ijwhether be white;

(7) if white is returned to (5);

(8) if not white judges Y _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}whether be white entirely;

(9) if be white entirely, return to (5);

(10) if not white, j=j+1 reads next line, judgement X _ijcolor;

(11) if white is returned to (5);

(12) if grey is returned to (5);

(13) if black judges Y _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}whether be white entirely;

(14) if not white is returned to (5);

(15) if be white entirely, fragment i and i ' matching process finish, and fragment i and i ' do not mate;

(16) if j+1 exceeds line range, fragment i and i ' matching process finish, fragment i and i ' coupling.

Ii. according to the criterion of step I, determine the mathematical model of images match index, be specially:

$S_{{\mathrm{ii}}^{\&prime;}}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{T}_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)$

Wherein:

S _{ii '}: i opens the match index of fragment and i ' fragment;

N: the sum of pixel on the vertical height of a scrap of paper;

X _ij: the gray-scale value of the capable pixel of fragment i right margin j;

T _{i '}(X _ij): judgement i opens right margin feature that shredded paper j is capable and opens the match index of the left margin feature of paper slip with the i ' of corresponding row;

This match index is specifically expressed as:

$T_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}{T}_1\left(X_{\mathrm{ij}}\right),& X_{\mathrm{ij}}=0,\\ T_2\left(X_{\mathrm{ij}}\right),& 0<X_{\mathrm{ij}}\&le;255.\end{array}\right.$

Wherein, t ₂(X _ij)=0.

And if only if S _{ii '}index is 0 o'clock, and two fragments are just considered as mating; If not 0 can not mate, and numerical value is larger, and matching degree is poorer.

5.2 steps 5.1 have completed the local matching process between two fragments, and a scrap of paper that meets matching condition that step 5.1 is obtained, forms each little fragment collection, to the capable coupling of fragment collection and row coupling, and the detailed process of the capable coupling of i.:

I) take a wherein fragment is benchmark, if an a scrap of paper is merged into two fragments in the success of the local matching of two fragments, puts into new fragment collection; If local matching is unsuccessful, retain benchmark fragment, continue local matching.When the concentrated fragment of primary fragment all cannot successful local matching, all put into new fragment collection;

Ii) new fragment repeats according to above-mentioned steps, until all a scrap of papers are spliced into complete row of tiles.

Ii. according to said process, determine the mathematical model of image line match index, be specially:

Objective function:

W＝min∑S _ii′

Constraint condition:

$\left\{\begin{array}{c}{S}_{{\mathrm{ii}}^{\&prime;}}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{T}_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)\\ T_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}{T}_1\left(X_{\mathrm{ij}}\right),& X_{\mathrm{ij}}=0,\\ T_2\left(X_{\mathrm{ij}}\right),& 0<X_{\mathrm{ij}}\&le;255.\end{array}\right.\\ a\&GreaterEqual;M\end{array}\right.$

Wherein, a is S _{ii '}number;

The minimum value of W is 0;

M is the longitudinally number of times of cutting of a scrap of paper

Iii forms row of tiles by the fragment collection mating by row, row of tiles matrix is carried out to transposition, then be listed as coupling with same method;

6, the image after step 5 coupling is reduced.

Effective interests of the present invention are:

Can disposable splicing process the comparatively huge a scrap of paper of quantity, and just coupling splicing flow process has proposed corresponding optimal solution, it is mainly reflected in:

(1) for Chinese fragment, only characteristic processing is done in border, boundary characteristic is carried out to Mathematical Models, therefore, the present invention is when upgrading image pattern database, and the scope of scan database is dwindled greatly, the in the situation that of a large amount of shredded paper to be spliced, there is jump.

(2) the designed edge following algorithm of the present invention, can guarantee the uniqueness in shredded paper matching process, has further improved validity of the present invention and operability.

Accompanying drawing explanation

Fig. 1 is a scrap of paper word perspective view of the embodiment of the present invention;

Fig. 2 is the local matching figure of the embodiment of the present invention;

Fig. 3 is the edge following algorithm process flow diagram of the embodiment of the present invention.

Embodiment

Chinese one side take below as example, implementation of the present invention is described simply.This example has been selected 209 a scrap of paper images altogether, an A4 paper of this 209 fragments crosscut 10 cuttves, rip cutting 18 cuttves.Concrete execution step is as follows:

(1) pre-service

(a) carry out preparation and the pre-service of image data set, comprise image digitazation, denoising, binaryzation;

(b) a scrap of paper is arranged as to the matrix of 11 row 19 row, by order from left to right, from top to bottom successively from 1 to 209 numbering;

(2) extract the feature set of fragment matrix, these features comprise: the position of the outermost pixel of a scrap of paper and gray-scale value, coboundary height;

(3) utilize eigenwert to classify to fragment:

1. according to sheet edge, whether there is strokes of characters information, shredded paper is divided three classes: 149 of each 19 of border, left and right fragment each 11, up-and-down boundary fragment and middle fragments;

2. according to the coboundary height extracting, fragment is classified:

Generally, white coboundary height or the black coboundary height with row of tiles is roughly the same.Calculate the boundary height of each fragment and the fragment with identical boundary height is classified as to a class, and adding up such number of tiles.Statistics as shown in Table 1.

Table one has the number of tiles of identical coboundary height

White top and dark top one have 43 groups as can be seen from Table I, and picture is only cut into for 11 row, therefore, need to do further processing to the classification of having divided.

Divide classification and need to follow certain condition:

(6) number of tiles of each class must be equal to or slightly less than 19;

(7), with other classifications highly separately, if quantity is less than 10, be not independently a classification (for example height 3);

(8) continuous several classifications are classified as same class highly mutually;

(9) final classification adds up to 11 classes;

(10) if still cannot determine classification, according to bottom level, carry out auxiliary judgment again.

Through further processing classification situation later as shown in Table 2:

The classification of table two coboundary height and corresponding number of tiles

(4) based on edge following algorithm, fragment is carried out to row, column coupling;

(5) image after display splicing:

The spliced fragment number table of table three

049	054	065	143	186	002	057	192	178	118	190	095	011	022	129	028	091	188	141
																			061	019	078	067	069	099	162	096	131	079	063	116	163	072	006	177	020	052	036
168	100	076	062	142	030	041	023	147	191	050	179	120	086	195	026	001	087	018
																			038	148	046	161	024	035	081	189	122	103	130	193	088	167	025	008	009	105	074
071	156	083	132	200	017	080	033	202	198	015	133	170	205	085	152	165	027	060
																			014	128	003	159	082	199	135	012	073	160	203	169	134	039	031	051	107	115	176
094	034	084	183	090	047	121	042	124	144	077	112	149	097	136	164	127	058	043
																			125	013	182	109	197	016	184	110	187	066	106	150	021	173	157	181	204	139	145
029	064	111	201	005	092	180	048	037	075	055	044	206	010	104	098	172	171	059
																			007	208	138	158	126	068	175	045	174	000	137	053	056	093	153	070	166	032	196
089	146	102	154	114	040	151	207	155	140	185	108	117	004	101	113	194	119	114

Claims (10)

1. a method for regular shredded paper automatic Mosaic, is characterized in that: the method mainly comprises following six steps:

(1) preparation of image data set and pre-service, comprise to shredded paper be numbered, image digitazation, image value, denoising point;

(2) shredded paper integral body is classified, by Chinese and English, single two-sided 4 kinds of situations that are divided into: Chinese one side, two-sided, the English one side of Chinese, English two-sided;

(3) extract respectively the feature of every width image local area, these features of Chinese are comprised: the position of a scrap of paper boundary pixel point and gray-scale value, up-and-down boundary height; Extraction scope to English a scrap of paper feature expands, and supplementary features also comprise: the row of English a scrap of paper is high, the horizontal level of English a scrap of paper is, the line space of English a scrap of paper;

(4) feature set of extracting according to step 3, classifies to fragment: be first divided into up-and-down boundary fragment, border, left and right fragment, middle fragment three classes; According to line space feature, up-and-down boundary height, horizontal level feature, the fragment with same characteristic features is divided into a class again;

(5) a: according to edge following algorithm, fragment is carried out to local matching, determine the match index of fragment boundary characteristic; B: a scrap of paper that meets matching condition that step a is obtained, utilizes the capable coupling of edge following algorithm and row coupling;

(6) image after step 5 coupling is reduced.

2. the method for a regular shredded paper automatic Mosaic according to claim 1, is characterized in that: the described method for numbering serial of step (1) is: by a scrap of paper from left to right, number consecutively from top to bottom, be designated as 1,2,3n; If desired distinguish pros and cons, front is designated as a1, a2, a3an; Reverse side is designated as b1, b2, b3bn; The described image digitazation of step (1) refers to: using pixel as least unit, and extract gray-scale value and the position of each pixel, set up Jacobian matrix; The described image value of step (1) refers to: the point that gray-scale value is " 0 " is black color dots, and the point that gray-scale value is " 255 " is white point, and between " 0 " and " 255 " is Grey Point; The described denoising point of step (1) refers to: for same color dot, surround heterochromatic point completely, by the same color around that changes into of the color of heterochromatic point.

3. the method for a regular shredded paper automatic Mosaic according to claim 1, it is characterized in that: the position of a scrap of paper boundary pixel point that step (3) is described and grey value characteristics and extracting method are: defining the most left or low order end one row pixel of a scrap of paper is left or right border, top or lowermost end one-row pixels point are up-and-down boundary, extract position and the gray-scale value of each boundary pixel point;

Described up-and-down boundary altitude feature and the extracting method of step (3) is: according to the whether complete white of the up-and-down boundary of each fragment, be divided into white border height and the large class of black border height two; Concrete sorting technique is as follows:

The lowermost end of fragment of take is x axle, take the fragment left side perpendicular to x axially upper be y axle, the joining of x axle and y axle is that initial point is set up coordinate system, and each pixel on picture is done to transverse projection to y axle; The projection of black or Grey Point is designated as effective projection, and projection number of times adds 1, and the projection of white point is invalid, and projection number of times does not change; Recording distance initial point is the projection number of times f (h) on the subpoint of h pixel;

$g\left(h\right)=\left\{\begin{array}{c}0,f\left(h\right)<n/10\\ 255,f\left(h\right)\&GreaterEqual;n/10\end{array}\right.$

When projection number of times f (h) be less than the total pixel number n of this row 1/10 time, the gray-scale value g (h) that puts h on y axle is designated as to " 0 "; When projection number of times f (h) be more than or equal to the total pixel number n of this row 1/10 time, the gray-scale value g (h) of a h is designated as to " 255 ";

On axis of projection, from the coboundary of fragment, add up downwards successively, until there is the point that color is different; This section is highly coboundary height, and lower boundary height is as the same;

The high feature of row and the extracting method of the English a scrap of paper that step (3) is described are: the height of English alphabet and in same a line shared position height roughly the same, therefore, English fragment is carried out after transverse projection, the interval that gray-scale value is " 1 " is alphabetical valid interval, the height of definition valid interval, for row is high, extracts row high;

Horizontal level feature and the extracting method of the English fragment that step (3) is described are: English fragment is carried out after transverse projection, up-and-down boundary between effective projection in zone of letter, the place horizontal level that is called this row letter apart from the distance at a scrap of paper top, extracts horizontal level feature;

Line space feature and the extracting method of the English a scrap of paper that step (3) is described are: the vertical range of two horizontal levels is defined as line space, extract line space feature.

4. the method for a regular shredded paper automatic Mosaic according to claim 1, is characterized in that: the feature set of extracting according to step 3 that step (4) is described, fragment is classified: concrete steps are:

I) according to scraps of paper border, whether there is strokes of characters information, shredded paper is divided three classes: up-and-down boundary fragment, border, left and right fragment and middle fragment;

Ii) according to line space feature, respectively above-mentioned three class a scrap of papers further to be classified, identical line space is divided into a class;

Iii) according to up-and-down boundary height, the formed three class fragment collection of step I to be classified, the fragment that up-and-down boundary height is identical or close is divided into same fragment collection;

Iv) utilize horizontal level feature, formed each fragment collection of step I i is further classified, the fragment in same level position is divided into a class.

5. according to the method for a regular shredded paper automatic Mosaic described in claim 1 or 4, it is characterized in that: described step I ii), dividing classification is that the condition that need to observe is:

(1) number of tiles of each class must be equal to or slightly less than the rip cutting number of times of paper;

(2) with other classifications highly separately, if quantity is less than 1/5 of number of tiles of all categories, be not independently a classification;

(3) continuous several classifications are classified as same class highly mutually;

(4) final classification adds up to the crosscut number of times of paper;

(5) if still cannot determine classification, according to bottom level, carry out auxiliary judgment again.

6. the method for a regular shredded paper automatic Mosaic according to claim 1, is characterized in that: the described step that fragment is mated of step (5) comprises:

A, according to edge following algorithm, fragment being carried out to local matching, is the coupling between two fragments, determines the match index of fragment boundary characteristic;

B, a scrap of paper that meets matching condition that step a is obtained, form each little fragment collection, utilizes edge following algorithm to the capable coupling of fragment collection and row coupling.

7. the method for a regular shredded paper automatic Mosaic according to claim 6, is characterized in that: the criterion of the local matching described in step a is: definition X _ijbe the gray-scale value that i opens the capable pixel of j on fragment right margin, definition Y _{i ' j}it is the gray-scale value (i ≠ i ') of the capable pixel of j on i ' fragment left margin; Judge that the key of whether mating is X _ijand Y _{i ' j}between matching degree, the feature set that step 3 is extracted, is characterized as benchmark with right margin, definition criterion is:

X _ijfor white, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}occur greyish white black three looks and entirely for black for normal, can mate;

X _ijfor grey, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}occur that random colour is normally, can mate;

X _ijfor black, Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}be not that white is normal entirely, can mate;

All the other situations are undesired, can not mate;

X _ijwith Y _{i ' j-1}, Y _{i ' j}, Y _{i ' j+1}relation as shown in Table 1;

Wherein: X _ij: i opens the gray-scale value of the capable pixel of j of Far Left one row of paper slip;

Y _{i ' j}: the gray-scale value of the capable pixel of j of rightmost one row of i ' paper slip.

8. a kind of regular shredded paper method for automatically split-jointing according to claim 7, is characterized in that: according to the criterion of described local matching, determine the mathematical model of images match index, be specially:

$S_{{\mathrm{ii}}^{\&prime;}}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{T}_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)$

Wherein:

S _{ii '}: i opens fragment and i ' the match index that fragment is total;

N: the sum of pixel in a scrap of paper vertical height;

X _ij: the gray-scale value of the capable pixel of fragment i right margin j;

T _{i '}(X _ij): the i ' that judgement i opens right margin feature that shredded paper j is capable and corresponding row open paper slip left margin feature the match index of data;

This match index is specifically expressed as:

$T_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}{T}_1\left(X_{\mathrm{ij}}\right),& X_{\mathrm{ij}}=0,\\ T_2\left(X_{\mathrm{ij}}\right),& 0<X_{\mathrm{ij}}\&le;255.\end{array}\right.$

Wherein, t ₂(X _ij)=0

And if only if S _{ii '}index is 0 o'clock, and two fragments are just considered as mating; If not 0 can not mate, and numerical value is larger, and matching degree is poorer.

9. the method for a regular shredded paper automatic Mosaic according to claim 6, is characterized in that: the edge following algorithm idiographic flow described in step a is as follows:

1) choose fragment i and i ';

2) suppose that fragment i and i ' mate mutually;

3) read the capable pixel X of fragment i right margin j _ijgray-scale value;

4) surface sweeping fragment i ' left margin and X _ijadjacent j-1, j, the pixel Y that j+1 is capable _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}, judge whether it is white entirely;

5) if be white entirely, and exceed line range, after j+1, return to 3);

6) if be not white entirely, j+1 reads next line, judgement X _ijwhether be white;

7) if white returns to 5);

8) if not white judges Y _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}whether be white entirely;

9) if be white entirely, return to 5);

10) if not white, j+1 reads next line, judgement X _ijcolor;

11) if white returns to 5);

12) if grey returns to 5);

13) if black judges Y _{i ' (j-1)}, Y _{i ' j}, Y _{i ' (j+1)}whether be white entirely;

14) if not white returns to 5);

15) if be white entirely, fragment i and i ' matching process finish, and fragment i and i ' do not mate;

16) if j+1 exceeds line range, fragment i and i ' matching process finish, fragment i and i ' coupling.

10. the method for a regular shredded paper automatic Mosaic according to claim 6, is characterized in that: the going detailed process of coupling and row coupling described in step b is:

(1) detailed process of row coupling:

I) take a wherein fragment is benchmark, if an a scrap of paper is merged into two fragments in the success of the local matching of two fragments, puts into new fragment collection; If local matching is unsuccessful, retain benchmark fragment, continue local matching; When the concentrated fragment of primary fragment all cannot successful local matching, all put into new fragment collection;

Ii) new fragment repeats according to above-mentioned steps, until all a scrap of papers are spliced into full line;

(2) according to said process, determine the mathematical model of image line match index, be specially:

Objective function:

W＝min∑S _ii′

Constraint condition:

$\left\{\begin{array}{c}{S}_{{\mathrm{ii}}^{\&prime;}}=\underset{j=1}{\overset{N}{\mathrm{\&Sigma;}}}{T}_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)\\ T_{i^{\&prime;}}\left(X_{\mathrm{ij}}\right)=\left\{\begin{array}{cc}{T}_1\left(X_{\mathrm{ij}}\right),& X_{\mathrm{ij}}=0,\\ T_2\left(X_{\mathrm{ij}}\right),& 0<X_{\mathrm{ij}}\&le;255.\end{array}\right.\\ a\&GreaterEqual;M\end{array}\right.$

Wherein, a is S _{ii '}number;

The minimum value of W is 0;

M is the number of times of a scrap of paper transverse cuts;

(3) the fragment collection mating by row is formed to row of tiles, row of tiles matrix is carried out to transposition, then be listed as coupling, the detailed process of row coupling is identical with row coupling.