CN104751177A - Method for identifying given blurred digital straight-line segment in digital image - Google Patents
Method for identifying given blurred digital straight-line segment in digital image Download PDFInfo
- Publication number
- CN104751177A CN104751177A CN201510137710.3A CN201510137710A CN104751177A CN 104751177 A CN104751177 A CN 104751177A CN 201510137710 A CN201510137710 A CN 201510137710A CN 104751177 A CN104751177 A CN 104751177A
- Authority
- CN
- China
- Prior art keywords
- bdss
- line segment
- coordinate
- straight
- pixel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a method for identifying a given blurred digital straight-line segment in a digital image. The method includes that when a blurred digital straight-line segment (BDSS) in the digital image is identified by a BDSS recognition algorithm, the BDSS is stored to be an identification result in a set form; geometrical linear characteristics of the BDSS given by the recognition result, and a rectangular frame with the slope meeting the geometrical linear characteristics is utilized to identify the given pixel set. The method is especially applicable to visual identification of the digital straight-line segment, digital images like the BDSS have the pixel set with the straight-line geometrical characteristics, no other original information in the digital images is modified except the rectangular frame added in the digital images, and the background is easy to observe. The method is simple in implementation, time consumption and resolution are increased linearly in a stabilizing mode, and the method is high in efficiency.
Description
Technical field
The present invention relates to a kind of method for fuzzy digit straight-line segment given in reference numbers image, namely by the set disperseed or contiguous pixels forms, carry out a kind of computing method of visual representation.
Background technology
In Digital Image Processing, the object in recognition image with particular community has very high science and using value, such as recognition of face, optical character identification etc.Digital direct line segment and derivative, the figure identification with rectilinear geometry feature belongs to the important branch in field of image recognition, it applies the road Identification (S.Aliana comprised in satellite photo, V.A.Tolpekina, W.Bijkera, L.Kumarb, " Identifyingcurvature of overpass mountain roads in Iran from high spatial resolutionremote sensing data ", Int.J.of Appl.Earth Observation andGeoinformation, vol.26, pp.21 – 25, 2014.), weld seam recognition (L.Jia in welding, N.Sun, " A line segment detection algorithm based on statisticalanalyses of quantified directions in digital image ", Comput.Modelling & New Technologies, vol.18, no.6, pp.79-88, 2014.) etc.For digital direct line segment and derivative object automatic identification computing method at home and abroad academia constantly weed out the old and bring forth the new, although accuracy of identification and efficiency progressively rise, adopt more coarse method mark result, as accompanying drawing illustrates shown in middle Fig. 1 always.Fig. 1 illustrates three kinds of common identification methods, be followed successively by Endpoint ID method (P.Bhowmickand B.B.Bhattacharya from left to right, " Fast polygonal approximation of digital curvesusing relaxed straightness properties ", IEEE Trans.Pattern Anal.Mach.Intell., vol.29, no.9, pp.1590-1602, Sept.2007.), multi-color cord segment identification method (L.Jia, N.Sun, " A line segment detection algorithm based onstatistical analyses of quantified directions in digital image ", Comput.Modelling & New Technologies, vol.18, no.6, pp.79-88, 2014.) and slope covering marking method (L.Buzer, " A simple algorithm for digital linerecognition in the general case ", Pattern Recognition, vol.40, no.6, pp:1675-1684, Jun.2007.).The common drawback of these three kinds of label methods uses monochromatic covering marking object completely, too much have modified original image data, background information is caused to be lost, mark overlap-add region cannot be known, object is identified to eye-observation and causes obstacle, particularly when identifying that object only comprises the partial pixel of overlay area, the situation that this object comprises partial pixel is the common situations in real world images.Academia often carries out test with the corresponding original image of (a) in similar Fig. 1 and (c), ideally this, the problem that monochromatic covering causes is not remarkable, but when relating to practical application, often need repeatedly to carry out experiment according to real image test result, three kinds of label methods similar to Figure 1 can cause the loss completely of identified areas background information and be difficult to distinguish overlap-add region in actual experiment, make test result be difficult to assess with human eye.Therefore, for figure identification, the particularly identification of digital direct line segment and derivative object, lacks one and had both implemented simply, again can the mark computing method of relatively intact preservation image background.
Summary of the invention
The technical problem to be solved in the present invention is: in order to overcome existing digital direct line segment and derivative object automatic identification after label method use monochromatic covering marking object completely, too much have modified original image data, background information is caused to be lost, mark overlap-add region cannot be known, the deficiency that object causes obstacle is identified to eye-observation, the invention provides a kind of method for fuzzy digit straight-line segment given in reference numbers image, be specially adapted to visual representation digital direct line segment, there is in the digital pictures such as fuzzy digit straight-line segment the pixel set of rectilinear geometry feature, except adding except rectangular shaped rim in digital picture, do not revise other raw informations in digital picture.
For making statement cheer and bright, the partial symbols that existing centralized definition is involved in the present invention and concept.
Z
+represent Positive Integer Set.
Z represents the integer set comprising zero.
R
+represent the arithmetic number set comprising zero.
R represents the real number set comprising zero.
Max{ element | the condition that element meets } represent " maximum " element satisfied condition.
Min{ element | the condition that element meets } represent " minimum " element satisfied condition.
For r ∈ R
+,
representative digit image wide.
the height of representative digit image.
There is r
1, r
2∈ R
+, satisfy condition
Index value:
View data: identify with index value and the digital information preserved.
Image space (is denoted as
): the view data of preserving digital picture.The unique corresponding pixel of each record cell.In view data, the pixel index value in the correspondence image information upper left corner is
according to correspondence image information, from left to right first, after order arrangement pixel from top to bottom, the kth ∈ Z in order
+the index value of individual pixel is
P
irepresent that in image space, index value is the pixel of i.
P
(x, y)represent in image space, the pixel of the point (x, y) in respective coordinates space, wherein x, y ∈ Z.
Dist (o
1, o
2): work as o
1, o
2during for point in two-dimensional space, dist (o
1, o
2) represent o
1with o
2between Euclidean distance; Work as o
1, o
2when being respectively point in two-dimensional space and straight line, dist (o
1, o
2) represent from o
1do and o
2vertical straight line, vertical line and o
2intersection point to o
1euclidean distance; Work as o
1, o
2∈ R, dist (o
1, o
2)=| o
1-o
2|.
For r, round (r)=min{z ∈ Z
+| dist (z, r)>=0}.
Side-play amount: for
side-play amount is
Serializing: will
in the order that increases progressively by index value of pixel, arrangement in a row one by one to the right, when number of pixel per line is
then newline immediately below this journey first pixel, and the aligning method pressing lastrow continues arrangement pixel, so repeatedly, until
in the order arrangement that increases progressively by index value of all pixels complete, now number of pels per line is
can regard as and there is length
wide
and unit Euclidean distance is
two-dimensional space.
Coordinate space (is denoted as
): serializing
and optionally one of them pixel is the rectangular coordinate system of initial point.For the two point (x in this coordinate system with unit Euclidean distance
1, y
1) and (x
2, y
2), by formula
the unit Euclidean distance values provided is
this coordinate system is the two-dimensional space with endless (y-axis) wide (x-axis).
Mean center: for
wherein i=1,2 ... n, n be sum of all pixels contained by BDSS; Mean center
coordinate be
Picture centre: i.e. pixel
Coordinate centralization: serializing
if
initial point for picture centre and x positive axis points to the picture centre right side of being expert at, d is known as side-play amount
ipixel
with
middle coordinate is
some one_to_one corresponding, (x
i, y
i) be called p
icentre coordinate.Work as pixel
centre coordinate be known as (x
i, y
i) time, its side-play amount is
index value is
Digital direct line segment (digital straight segment, DSS): coordinate centralization
if pixel
with
coordinate figure be respectively
with
i
p, i
q, j
p, j
q∈ Z and i
p< i
q, the straight line crossing p and q is denoted as
and its slope belongs to interval (0,1), for
cathetus x=i ∈ R wherein i
p≤ i≤i
qif, straight line
be (x, y) with the intersecting point coordinate of straight line x=i, then there is unique pixel that coordinate figure is (round (x), round (y)), the Euclidean distance of itself and intersection point is not more than
at interval [i
p, i
q] interior change i value, can obtain the set of this type of pixel, digital direct line segment DSS (p, q) determined by p and q is:
Fuzzy digit straight-line segment (blurred digital straight segment, BDSS): coordinate centralization
if i, j, a, b, μ, ω ∈ is Z, b ≠ 0 and gcd (a, b)=1, the then μ≤di-bj that satisfies condition≤μ+ω's
intersection be called fuzzy digit straight-line segment BDSS (a, b, μ, ω), that is:
On basis defined above, the technical solution adopted for the present invention to solve the technical problems is:
When fuzzy digit straight-line segment BDSS recognizer completes wide
high
digital picture in the identification of fuzzy digit straight-line segment BDSS, fuzzy digit straight-line segment BDSS saves as recognition result with the form of set;
According to the fuzzy digit straight-line segment BDSS that recognition result is given, calculate its geometry linear characteristic, the rectangular shaped rim using slope to meet this geometry linear characteristic identifies given pixel set.
Specifically comprise the following steps:
Step 1: in coordinates computed centralization image space, respective coordinates space
centre coordinate;
Step 2: according to the centre coordinate of pixel in fuzzy digit straight-line segment BDSS, calculates mean center, and uses linear regression to calculate the direction θ of fuzzy digit straight-line segment BDSS;
Step 3: the converted coordinate calculating pixel in fuzzy digit straight-line segment BDSS: by coordinate space corresponding for coordinate centralization
origin translation to mean center, and x-axis positive axis to be rotated to overlapping with θ;
Step 4: calculate fuzzy digit straight-line segment BDSS in converted coordinate system according to the converted coordinate of pixel in fuzzy digit straight-line segment BDSS
in four frontier points;
Step 5: generate the boundary line l being parallel to direction θ
1and l
2, and calculate its converted coordinate, boundary line l
1and l
2respectively by wherein two frontier points in step 4;
Step 6: generate the boundary line l perpendicular to direction θ
3and l
4, and calculate its converted coordinate, boundary line l
3and l
4respectively by two other frontier point in step 4, straight line l
1, l
2, l
3and l
4intersect and form rectangular shaped rim;
Step 7: the converted coordinate generated according to step 5 and 6, calculates its centre coordinate;
Step 8: the centre coordinate generated according to step 7, calculates its side-play amount and side-play amount is saved as result output.
According to side-play amount determine that the set of pixel is rectangular shaped rim, finally, pixel amending method can be used, identify BDSS digital picture in change the value of pixel contained by rectangular shaped rim to identify fuzzy digit straight-line segment BDSS.
In step 1 and 2, will identify that the digital picture of fuzzy digit straight-line segment BDSS is mapped as
coordinate centralization
for any pixel p ∈ BDSS, if the side-play amount of p is d, then its centre coordinate calculates according to formula (1):
By the centre coordinate of BDSS pixel, calculate the mean center of BDSS
According to linear regression theory, the direction θ ∈ of BDSS [0, π), calculated by formula (2):
wherein i=1,2 ... n, n be sum of all pixels contained by BDSS.
In step 3, for
its converted coordinate
calculate according to formula (3):
In step 4, converted coordinate system
in four frontier points obtained by all horizontal ordinate x and ordinate y comparing converted coordinate.
Coordinate space
the converted coordinate of initial point be
in step 7, for
its centre coordinate
calculate according to formula (7):
In step 8, for
the side-play amount d of pixel is calculated by formula (8):
?
in, because initial point is positioned at the mean center of BDSS, there are four extreme values in BDSS pixel transitions coordinate, i.e. horizontal ordinate maximal value x '
lmax, horizontal ordinate minimum value x '
lmin, ordinate maximal value y '
wmaxwith ordinate minimum value y '
wmin;
in also exist corresponding conversion coordinate (x '
lmax, 0), (x '
lmin, 0), (0, y '
wmax) and (0, y '
wmin) pixel, these pixels are the frontier point described in step 4;
Wherein,
represent rectangular shaped rim, rectangular shaped rim
definition provided by formula (4):
Wherein
m=1,2,3,4,
represent from p '
(x ', y ')do and l
mvertical straight line, vertical line and l
mintersection point to p '
(x ', y ')euclidean distance.
Rectangular shaped rim
in fact be made up of four line segments, formula (4) can be replaced by formula (5) and formula (6), and formula (5) and formula (6) are respectively l
1and l
2on, l
3and l
5on
the converted coordinate of pixel:
The invention has the beneficial effects as follows, a kind of method for fuzzy digit straight-line segment given in reference numbers image of the present invention, to given digital direct line segment, fuzzy digit straight-line segment etc., by the pixel set disperseed or contiguous pixels forms, calculate its geometry linear characteristic, the rectangular shaped rim using slope to meet institute's estimated performance identifies the computing method of given pixel set, and tool has the following advantages:
(1) the inventive method is under the prerequisite significantly not revising background pixel, clearly designates given fuzzy digit straight-line segment.Along with the rising of identified digital image resolution, the inventive method revise pixel quantity decline fast with the ratio of the whole pixel of image, its visual effect is rendered as height respectively in rate image, and the BDSS background of the present invention's mark is more easily observed; Even little image in different resolution, the inventive method is also more limited to taking of image background.
(2) the inventive method is implemented simple, and consuming time substantially linearly stablizing with resolution increases, performance efficiency.
Accompanying drawing explanation
Below in conjunction with drawings and Examples, the present invention is further described.
Fig. 1 is identification method conventional at present.(a) Endpoint ID method, (b) multi-color cord segment identification method, (c) slope covering marking method.
Fig. 2 is the schematic diagram of a kind of method for fuzzy digit straight-line segment given in reference numbers image of the present invention.A former coordinate system is rotated to converted coordinate system by (), converted coordinate system is rotated back to former coordinate system by (b).
Fig. 3 is a kind of data structure for Point class in the method for fuzzy digit straight-line segment given in reference numbers image of the present invention.
Fig. 4 is the theory diagram of a kind of method for fuzzy digit straight-line segment given in reference numbers image of the present invention.
Fig. 5 is a kind of theory diagram for step 2 in the method for fuzzy digit straight-line segment given in reference numbers image of the present invention.
Fig. 6 is a kind of theory diagram for step 5 in the method for fuzzy digit straight-line segment given in reference numbers image of the present invention.
Fig. 7 is a kind of theory diagram for step 6 in the method for fuzzy digit straight-line segment given in reference numbers image of the present invention.
Fig. 8 is the result figure that the present invention identifies given fuzzy digit straight-line segment in the image of 128*128 resolution.(a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
Fig. 9 is the result figure that the present invention identifies given fuzzy digit straight-line segment in the image of 256*256 resolution.(a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
Figure 10 is the result figure that the present invention identifies given fuzzy digit straight-line segment in the image of 512*512 resolution.(a) image cameraman, (b) image house, (c) image lena, (d) image puzzle.
Figure 11 is that a kind of method different step for fuzzy digit straight-line segment given in reference numbers image of the present invention is at the situation analysis figure average consuming time completing all BDSS frames in drafting one width figure.
Figure 12 is that a kind of method different step for fuzzy digit straight-line segment given in reference numbers image of the present invention is at the situation analysis figure consuming time completing BDSS frame the longest in drafting one width figure.
Figure 13 is that a kind of method different step for fuzzy digit straight-line segment given in reference numbers image of the present invention is at the situation analysis figure consuming time completing BDSS frame the shortest in drafting one width figure.
Figure 14 is that a kind of method for fuzzy digit straight-line segment given in reference numbers image of the present invention is always consuming time to be increased and the situation analysis figure of change with image resolution ratio.
Embodiment
In conjunction with the accompanying drawings, the present invention is further detailed explanation.
When BDSS recognizer completes wide
high
digital picture in the identification of fuzzy digit straight-line segment (BDSS), BDSS saves as recognition result with the form of set.
For arbitrary BDSS, according to BDSS definition, shown in Fig. 2 (a) that the similar accompanying drawing of its form illustrates and Fig. 2 (b).Fig. 2 (a) and Fig. 2 (b) pixel group that square represents respectively show a BDSS, former coordinate system is rotated to converted coordinate system by Fig. 2 (a) expression, and converted coordinate system is rotated back to former coordinate system by Fig. 2 (b) expression.Target of the present invention is according to given BDSS, by the calculating based on mathematical principle, obtains as shown in Fig. 2 (a), by straight line l
1, l
2, l
3and l
4intersect the rectangular shaped rim of formation in the digital picture identifying this BDSS, the side-play amount of corresponding pixel.Every bar straight-line pass frontier point, and l
1, l
2parallel with BDSS direction θ, l
3, l
4vertical with θ.
According to side-play amount determine that the set of pixel is called rectangular shaped rim, the various pixel amending methods beyond the present invention can be used, identify BDSS digital picture in change the value of pixel contained by rectangular shaped rim to identify BDSS.
For given BDSS, will identify that the digital picture of BDSS is mapped as
coordinate centralization
the coordinate system that middle pixel center coordinate is corresponding
represented by orthogonal arrow straight line in fig. 2, the arrow indicating x represents x-axis positive axis.For any pixel p ∈ BDSS, if the side-play amount of p is d, then its centre coordinate calculates according to formula (1):
By the centre coordinate of BDSS pixel, the mean center of BDSS can be calculated
with solid point and character mark in Fig. 2
according to linear regression theory, the direction θ ∈ of BDSS [0, π) calculated by formula (2):
N is sum of all pixels contained by BDSS.
L in Fig. 2 (a)
1and l
2parallel, l
1and l
2with
the less angle of middle x-axis positive axis is θ.Though the centre coordinate according to θ and BDSS pixel directly can calculate l
1, l
2, l
3and l
4parameter and intersection point, but computation process is more complicated, and it is larger to implement difficulty.
The present invention has evaded direct calculating l
1, l
2, l
3and l
4the classic method of intersection point, the minimum and maximum coordinate figure of BDSS, completes the calculating of rectangular shaped rim by translation and rotating coordinate system and in calculating coordinate system after conversion.Translation be by
initial point move to the mean center of BDSS, rotation is rotated counterclockwise
x-axis positive axis, and ensure that the angle of postrotational x-axis positive axis and former x-axis positive axis is θ, by this translation with rotate the coordinate system obtained and be called converted coordinate system, be denoted as
in Fig. 2 (a),
represented by orthogonal arrow straight line, the arrow indicating x ' represents x-axis positive axis, and the arrow indicating y ' represents y-axis positive axis, curved arrow dotted line from
initial point points to
initial point
representing will
origin translation extremely
with
the angle of x-axis positive axis be θ, sense of rotation is represented by curved arrow solid line.BDSS pixel exists
in coordinate be called converted coordinate.For
its converted coordinate
calculate according to formula (3):
?
in, because initial point is positioned at the mean center of BDSS, there are four extreme values in BDSS pixel transitions coordinate, i.e. horizontal ordinate maximal value x '
lmax, horizontal ordinate minimum value x '
lmin, ordinate maximal value y '
wmaxwith ordinate minimum value y '
wmin.
in also exist corresponding conversion coordinate (x '
lmax, 0), (x '
lmin, 0), (0, y '
wmax) and (0, y '
wmin) pixel, these pixels are called frontier point, identify frontier point in Fig. 2 (a) with solid point and converted coordinate.
By frontier point, θ, l
1with l
2parallel, l
3with l
4parallel and l
1with l
3vertical condition, for Fig. 2 (a)
shown in situation, rectangular shaped rim
definition provided by formula (4):
Wherein d
m=dist (p '
(x ', y '), l
m), m=1,2,3,4, dist (p '
(x ', y '), l
m) represent from p '
(x ', y ')do and l
mvertical straight line, vertical line and l
mintersection point to p '
(x ', y ')euclidean distance.
For Fig. 2 (a)
shown in situation,
in fact be made up of four line segments, formula (4) can be substituted by formula (5) and formula (6):
Formula (5) and formula (6) sets forth l
1and l
2on, l
3and l
4on
the converted coordinate of pixel.Pass through
the converted coordinate of pixel, can calculate its pixel-shift amount, can be changed the value of respective pixel in the digital picture identifying BDSS by side-play amount.The process calculating side-play amount comprises successively according to converted coordinate computing center coordinate with according to centre coordinate calculating side-play amount.
According to converted coordinate computing center coordinate, be equivalent to by inciting somebody to action
origin translation return
original, and to turn clockwise
middle x-axis positive axis extremely
x-axis positive axis, will
be reduced to
in Fig. 2 (b) that accompanying drawing illustrates,
represented by orthogonal arrow straight line, the arrow indicating x ' represents x-axis positive axis, and the arrow indicating y ' represents y-axis positive axis,
represented by orthogonal arrow straight line, the arrow indicating x represents x-axis positive axis, from
initial point
point to
the curved arrow dotted line of initial point represents origin translation, from
x-axis positive axis point to
the curved arrow solid line of x-axis positive axis represent and be rotated counterclockwise.Reduction process needs
the converted coordinate of initial point, its converted coordinate is
for
its centre coordinate
calculate according to formula (7):
According to
the centre coordinate of pixel, for
the side-play amount d of pixel is calculated by formula (8):
Pass through
the side-play amount of pixel, can change in the digital picture that identification BDSS is used
the value of pixel, thus the object completing mark fuzzy digit straight-line segment BDSS.
Particularly, the embodiment of computing method of the present invention completes by writing computer program.
Program input comprises: a BDSS, identifies the digital picture that uses of BDSS and it is wide
and height
wherein the pixel of BDSS represents with the accompanying drawing data structure shown in Fig. 3 that is right.Fig. 3 illustrates Point class, its attribute comprises the Deviation for storing side-play amount, be respectively used to the attribute XtoImageCenter of storage center coordinate horizontal ordinate x and the attribute YtoImageCenter of ordinate y, be respectively used to store the attribute XtoCollectionCenter of converted coordinate horizontal ordinate x and the attribute YtoCollectionCenter of ordinate y.The data type of attribute Deviation is integer, represents with integer, and the data type of other Point generic attributes is double-precision floating points, represents with double.
It is calculate according to input BDSS that program exports, the pixel set defined by formula (4).
Program comprises eight steps, sees Fig. 4 that accompanying drawing illustrates.Now introduce eight steps successively.
Step 1: in coordinates computed centralization image space, the centre coordinate of the Point object in respective coordinates space.
Map digital picture to image space
according to
the index value of middle pixel, input
with
the side-play amount of each Point object can be calculated and coordinate centralization
for the Point object that side-play amount is d, its centre coordinate calculates according to formula (1), and horizontal ordinate x and the ordinate y of centre coordinate preserve with attribute XtoImageCenter and YtoImageCenter respectively.
Step 2: according to Point object coordinates in BDSS, calculates mean center, and uses linear regression to calculate its direction θ.
According to each Point object centers coordinate in the BDSS that step 1 calculates, calculate the mean center of BDSS
based on linear regression theory, calculate the direction θ of BDSS according to formula (2).Idiographic flow is shown in Fig. 6 that accompanying drawing illustrates.
Step 3: by the origin translation of coordinate space corresponding for coordinate centralization to mean center, and x-axis positive axis is rotated to overlapping with θ, calculate the converted coordinate of Point object in BDSS.
The converted coordinate of Point object calculates according to formula (3), and horizontal ordinate x and the ordinate y of converted coordinate preserve with attribute XtoCollectionCenter and YtoCollectionCenter respectively.
Step 4: according to the converted coordinate computation bound point of Point object in BDSS.
middle BDSS coordinate extreme value is that the XtoCollectionCenter property value and YtoCollectionCenter property value by comparing all Point object converted coordinates in BDSS obtains.According to comparing the extreme value obtained, generate similar accompanying drawing illustrate Fig. 2 in frontier point (x '
lmax, 0), (x '
lmax, 0), (0, y '
wmax) and (0, y '
wmax) Point object.These frontier points are bases that completing steps 5 and step 6 calculate.
Step 5: generate the Point object being parallel to the boundary line of direction θ, and calculate its converted coordinate.
L in corresponding diagram 2 is generated according to formula (5)
1and l
2point object, idiographic flow is shown in Fig. 6 that accompanying drawing illustrates.The coordinate of frontier point uses variable X respectively
start, X
end, Y
startand Y
endpreserve.Frontier point in the Fig. 2 illustrated with accompanying drawing (x '
lmax, 0), (x '
lmin, 0), (0, y '
wmax) and (0, y '
wmin) be example, X
start, X
end, Y
startand Y
endvalue be respectively x '
lmin, x '
lmax, y '
wminwith y '
wmax, i.e. the minimum value of horizontal ordinate x and maximal value, the minimum value of ordinate y and maximal value.At maintenance Y
startand Y
endconstant and X
starbe not more than X
endprerequisite under, with X
startfor converted coordinate horizontal ordinate x initial value and to circulate X at every turn
startfrom the mode of increasing 1, circulation T.G Grammar coordinate is respectively (X
start, Y
start) and (X
end, Y
start) two Point objects.All Point objects that circulation produces are kept in rectangular shaped rim.
Step 6: the Point object generating the boundary line perpendicular to direction θ, and calculate its converted coordinate.
L in corresponding diagram 2 is generated according to formula (6)
3and l
4point object, idiographic flow is shown in Fig. 7 that accompanying drawing illustrates.The minimum value of frontier point horizontal ordinate x is used to reinitialize variable X
start, at maintenance X
startand X
endconstant and Y
startbe not more than Y
endprerequisite under, with Y
startfor converted coordinate horizontal ordinate y initial value and to circulate Y at every turn
startfrom the mode of increasing 1, circulation T.G Grammar coordinate is respectively (X
start, Y
start) and (X
start, Y
end) two Point objects.All Point objects that circulation produces are kept in rectangular shaped rim.
Step 7: according to the converted coordinate of generated Point object, calculate its centre coordinate.
For Point objects all in rectangular shaped rim, the calculating of its centre coordinate completes according to formula (7).
Step 8: according to the centre coordinate of generated Point object, calculates its side-play amount and side-play amount is saved as result output.
For Point objects all in rectangular shaped rim, the calculating of its side-play amount completes according to formula (8).
According to Output rusults, pixel amending method can be used, in the digital picture identifying BDSS, change the value of pixel contained by rectangular shaped rim to identify fuzzy digit straight-line segment BDSS.
Below be divided into two to introduce beneficial effect of the present invention, Part I is the intuitive visual effect of the inventive method; Part II is counting yield analysis:
1, visual effect:
Fig. 8 to Figure 10 illustrates the result figure (in figure, rectangular shaped rim is all with the black line that 1 pixel is wide, white line, and the line segment of the laminated structure of black line represents) that application the inventive method identifies given BDSS.In Fig. 8 to Figure 10, the resolution of digital picture is followed successively by 128 × 128, and 256 × 256 and 512 × 512.After noticing that image resolution ratio is changed, the BDSS of input also can change, and the output of the inventive method can change thereupon, is not amplify identical result.Fig. 8 to Figure 10 improves for illustration of with resolution, the corresponding minimizing of background pixel that algorithm pollutes, and visual impression is that the background that frame blocks tails off, and more easily differentiates the BDSS of superposition mutually.
Result shown in result and Fig. 1 shown in comparison diagram 8 to Figure 10, be easy to show that the inventive method is under the prerequisite significantly not revising background pixel, clearly designate given BDSS, the image that the algorithm ID resolution rate of showing according to Fig. 1 (c) of can justing think is lower, such as Fig. 8, the rectangle that the BDSS then identified by the inventive method can fill with monochrome represent, thus causes human eye cannot identification BDSS background area.Along with the rising of identified digital image resolution, the inventive method revise pixel quantity decline fast with the ratio of the whole pixel of image, its visual effect is rendered as height respectively in rate image, the BDSS background of the present invention's mark is more easily observed, such as Figure 10, even little image in different resolution, the inventive method is also more limited to taking of image background, such as Fig. 8.
2, counting yield:
The histogram of Figure 11 to Figure 13 that accompanying drawing illustrates illustrates the time of eight steps consumption described in embodiment, notices that Figure 11 to Figure 13 longitudinal axis unit is the logarithm of cpu clock periodicity.First Figure 11 to Figure 13 transverse axis presses image cameraman, and the order of lena, house and puzzle from left to right arranges, and for each width figure, then the order of pressing rate 128 × 128,256 × 256 and 512 × 512 respectively from left to right arranges.For Figure 11, transverse axis comment " image cameraman " top column from left to right respectively correspondence respectively rate be 128 × 128,256 × 256 and 512 × 512 consuming time.Region representation different step consuming time of different colours in each column, region is larger, then this step is consuming time more, and column is higher, then always consuming time more.Same column in Figure 11 to Figure 13 each square from the bottom to top represents the time that the step 1 described in embodiment to each step of step 8 consumes successively.Figure 11 illustrates on average consuming time, namely to each step of mark and total mean value consuming time of all given BDSS in a width digital picture.Figure 12 illustrates maximum consuming time, namely to have in the BDSS mark of maximum Point object each step and always consuming time to given in a width digital picture.Figure 13 illustrates minimum consuming time, namely to have in the BDSS mark of minimum Point object each step and always consuming time to given in a width digital picture.Figure 11-13 is the distributions consuming time in varied situations of display different step, and for illustration of algorithm core procedure 5,6,7 is consuming time less.
By observing Figure 11 to Figure 13, step 8 shared region in multiple column is maximum, and namely this step is the most consuming time.This is because step 8 computation complexity is higher, and rectangular shaped rim will be converted to and meet subsequent treatment form and cause.For each Point object, 4 multiplication and 6 sub-addition computings will be completed according to formula (7) step 7, and step 8 need complete 1 multiplication, 2 divisions, 3 sub-additions and 2 rounding operations
although step 8 operation times is slightly little, mostly what it related to is division and rounds this type of comparatively complex calculations.Step 3 and step 4 is less than, particularly when BDSS comprises less Point object, namely shown in Figure 13 for calculating the step 5 of rectangular shaped rim and step 6 Figure 11 to Figure 13 is generally consuming time.This has reacted the inventive method from efficiency and has implemented simple feature, because step 5 and step 6 are the key steps producing rectangular shaped rim, because its implementation process does not relate to complex calculation or data structure, with other steps, particularly step 7 is compared with step 8, and step 5 is consuming time less with step 6.
Figure 14 always consuming timely increases and the situation analysis figure of change with image resolution ratio, is exactly process the summation that all BDSS are consuming time in piece image always consuming time.Figure 14 illustrates from rate 128 × 128 to 512 × 512 respectively, the total situation consuming time of the computing method of the present invention at interval of 32 × 32, notices that the transverse axis of Figure 14 illustrate only the one dimension size of image.By the above-mentioned order of rate respectively, namely 128,160,192 ... image cameraman, the consuming time of lena, house and puzzle different resolution version marks in fig. 14 with black circle, for identical content but the round dot of difference rate image respectively connect with the straight-line segment with same line section shape, concrete line segment shape illustrates and refers to Figure 14.Four lines in Figure 14 substantially linearly, namely to be consuming timely substantially directly proportional to resolution, the wherein distribution consuming time of typically image cameraman, although the distribution consuming time of image lena and image puzzle rises and falls very large, but from the angle analysis of linear regression, its distribution is less with the difference of ideal line.
In Figure 14, the distribution consuming time of image house increases with resolution, occurs downward trend, but does not all occur this trend due to other image, and the situation of house is only some phenomena.To sum up, by the distribution consuming time of Figure 14, do not find that the inventive method increases with resolution, obviously present the situation of non-linear quick increase consuming time, that is, the inventive method is consuming time increases with resolution is substantially linearly stable.This illustrates that the performance of computing method of the present invention is comparatively efficient.
Claims (8)
1., for a method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that,
When fuzzy digit straight-line segment BDSS recognizer completes wide
high
digital picture in the identification of fuzzy digit straight-line segment BDSS, fuzzy digit straight-line segment BDSS saves as recognition result with the form of pixel set;
According to the fuzzy digit straight-line segment BDSS that recognition result is given, calculate its geometry linear characteristic, the rectangular shaped rim using slope to meet this geometry linear characteristic identifies given pixel set.
2., as claimed in claim 1 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: specifically comprise the following steps:
Step 1: in coordinates computed centralization image space, respective coordinates space
centre coordinate;
Step 2: according to the centre coordinate of pixel in fuzzy digit straight-line segment BDSS, calculates mean center, and uses linear regression to calculate the direction θ of fuzzy digit straight-line segment BDSS;
Step 3: the converted coordinate calculating pixel in fuzzy digit straight-line segment BDSS: by coordinate space corresponding for coordinate centralization
origin translation to mean center, and x-axis positive axis to be rotated to overlapping with θ;
Step 4: calculate fuzzy digit straight-line segment BDSS in converted coordinate system according to the converted coordinate of pixel in fuzzy digit straight-line segment BDSS
in four frontier points;
Step 5: generate the boundary line l being parallel to direction θ
1and l
2, and calculate its converted coordinate, boundary line l
1and l
2respectively by wherein two frontier points in step 4;
Step 6: generate the boundary line l perpendicular to direction θ
3and l
4, and calculate its converted coordinate, boundary line l
3and l
4respectively by two other frontier point in step 4, straight line l
1, l
2, l
3and l
4intersect and form rectangular shaped rim;
Step 7: the converted coordinate generated according to step 5 and 6, calculates its centre coordinate;
Step 8: the centre coordinate generated according to step 7, calculates its side-play amount and side-play amount is saved as result output.
3. as claimed in claim 2 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: in step 1 and 2, will identify that the digital picture of fuzzy digit straight-line segment BDSS is mapped as
coordinate centralization
for any pixel p ∈ BDSS, if the side-play amount of p is d, then its centre coordinate calculates according to formula (1):
By the centre coordinate of BDSS pixel, calculate the mean center of BDSS
According to linear regression theory, the direction θ ∈ of BDSS [0, π), calculated by formula (2):
wherein i=1,2...n, n sum of all pixels contained by BDSS.
4., as claimed in claim 3 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: in step 3, for
its converted coordinate
calculate according to formula (3):
5., as claimed in claim 4 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: in step 4, converted coordinate system
in four frontier points obtained by all horizontal ordinate x and ordinate y comparing converted coordinate.
6., as claimed in claim 5 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: coordinate space
the converted coordinate of initial point be
in step 7, for
its centre coordinate
calculate according to formula (7):
7., as claimed in claim 6 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: in step 8, for
the side-play amount d of pixel is calculated by formula (8):
?
in, because initial point is positioned at the mean center of BDSS, there are four extreme values in BDSS pixel transitions coordinate, i.e. horizontal ordinate maximal value x '
lmax, horizontal ordinate minimum value x '
lmin, ordinate maximal value y '
wmaxwith ordinate minimum value y '
wmin;
in also exist corresponding conversion coordinate (x '
lmax, 0), (x '
lmin, 0), (0, y '
wmax) and (0, y '
wmin) pixel, these pixels are the frontier point described in step 4;
Wherein,
represent rectangular shaped rim, rectangular shaped rim
definition provided by formula (4):
Wherein d
m=dist (p '
(x ', y '), l
m), m=1,2,3,4, dist (p '
(x ', y '), l
m) represent from p '
(x ', y ')do and l
mvertical straight line, vertical line and l
mintersection point to p '
(x ', y ')euclidean distance.
8., as claimed in claim 7 for the method for fuzzy digit straight-line segment given in reference numbers image, it is characterized in that: formula (5) and formula (6) are respectively l
1and l
2on, l
3and l
4on
the converted coordinate of pixel:
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510137710.3A CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510137710.3A CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104751177A true CN104751177A (en) | 2015-07-01 |
| CN104751177B CN104751177B (en) | 2018-01-12 |
Family
ID=53590830
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510137710.3A Active CN104751177B (en) | 2015-03-26 | 2015-03-26 | A kind of method for being used in reference numbers image give fuzzy digit straightway |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104751177B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513044A (en) * | 2015-11-20 | 2016-04-20 | 常州大学 | Digital straight segment recognition method based on statistical measurement of straight line features |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4901365A (en) * | 1988-12-19 | 1990-02-13 | Ncr Corporation | Method of searching binary images to find search regions in which straight lines may be found |
| CN102156884A (en) * | 2011-04-25 | 2011-08-17 | 中国科学院自动化研究所 | Straight segment detecting and extracting method |
| CN102819743A (en) * | 2012-08-14 | 2012-12-12 | 常州大学 | Detection method for quickly identifying straight-line segments in digital image |
| CN103150741A (en) * | 2012-11-30 | 2013-06-12 | 常州大学 | Method for rapidly skeletonizing graph of binary digital image |
-
2015
- 2015-03-26 CN CN201510137710.3A patent/CN104751177B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4901365A (en) * | 1988-12-19 | 1990-02-13 | Ncr Corporation | Method of searching binary images to find search regions in which straight lines may be found |
| CN102156884A (en) * | 2011-04-25 | 2011-08-17 | 中国科学院自动化研究所 | Straight segment detecting and extracting method |
| CN102819743A (en) * | 2012-08-14 | 2012-12-12 | 常州大学 | Detection method for quickly identifying straight-line segments in digital image |
| CN103150741A (en) * | 2012-11-30 | 2013-06-12 | 常州大学 | Method for rapidly skeletonizing graph of binary digital image |
Non-Patent Citations (1)
| Title |
|---|
| 李沛等: "一种新的不规则直线段识别方法", 《光学技术》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513044A (en) * | 2015-11-20 | 2016-04-20 | 常州大学 | Digital straight segment recognition method based on statistical measurement of straight line features |
| CN105513044B (en) * | 2015-11-20 | 2018-07-17 | 常州大学 | A kind of digital direct straight line segments recognition method based on statistical measures linear feature |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104751177B (en) | 2018-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8780131B2 (en) | Systems and methods for text-based personalization of images | |
| Andres | Discrete linear objects in dimension n: the standard model | |
| CN103198479B (en) | Based on the SAR image segmentation method of semantic information classification | |
| CN111382658B (en) | Road traffic sign detection method in natural environment based on image gray gradient consistency | |
| CN107564087A (en) | A kind of Three-D linear symbol rendering intent based on screen | |
| US20200279415A1 (en) | Efficiently Computed Distance Fields | |
| CN106340010A (en) | Corner detection method based on second-order contour difference | |
| CN106408626A (en) | Graphics processing systems | |
| CN105068918A (en) | Page test method and device | |
| CN104949621B (en) | A kind of boundary alignment method of grating scale striped | |
| Shen et al. | A progressive method for the collapse of river representation considering geographical characteristics | |
| Li et al. | ColorSketch: A drawing assistant for generating color sketches from photos | |
| CN109712225B (en) | Rendering method, device, equipment and storage medium of surface object | |
| Gader | Fuzzy spatial relations based on fuzzy morphology | |
| CN112686872B (en) | Wood counting method based on deep learning | |
| CN108898679A (en) | A kind of method of component serial number automatic marking | |
| CN104751177A (en) | Method for identifying given blurred digital straight-line segment in digital image | |
| CN105512663A (en) | Significance detection method based on global and local contrast | |
| CN106056575B (en) | A kind of image matching method based on like physical property proposed algorithm | |
| Wang et al. | Improving facade parsing with vision transformers and line integration | |
| Bjørke | Exploration of information theoretic arguments for the limited amount of information in a map | |
| Kurlin et al. | A persistence-based approach to automatic detection of line segments in images | |
| Ren et al. | A new and fast contour-filling algorithm | |
| Ullah et al. | Experimental and numerical research of paved microcrack using histogram equalization for detection and segmentation | |
| CN102567946B (en) | Generation method of computer oil painting and system thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230414 Address after: Room 306, Block C, Tianrun Technology Building, Changzhou Science and Education City, No. 18-3 Changwu Middle Road, Wujin District, Changzhou City, Jiangsu Province, 213000 Patentee after: Changzhou Jingxun Micro Information Technology Co.,Ltd. Address before: Gehu Lake Road Wujin District 213164 Jiangsu city of Changzhou province No. 1 Patentee before: CHANGZHOU University |
|
| TR01 | Transfer of patent right |