Summary of the invention
To the problem that exists in the prior art; The object of the present invention is to provide a kind of characteristics information extraction method of static image target, be used to solve problem and the prior art that prior art is selected the image object characteristic information and acquiring method does not have a versatility and ask for the slow problem that characteristic information produces through the complicated calculations process.
For realizing above-mentioned purpose, the invention provides a kind of characteristics information extraction method of static image target, comprising:
Select to surround the rectangular area of said static image target;
Begin from border, said rectangular area, search for along level, vertical inside direction, obtain the frontier point of static image target by the SI of setting;
Confirm the rectangular area of static image target, the region of search of static image target and the RP of characteristic element according to said frontier point;
Ask for the characteristic element in the frontier point neighborhood; The coordinate of said characteristic element is the coordinate with pixel of maximum value factor; The optical parametric of said characteristic element is said R, G, B value with pixel of maximum value factor, and said maximum value factor comprises that 8 neighborhoods are worth factor or 16 neighborhoods are worth the maximum value factor in the factor;
RP by the rectangular area of a stack features element, static image target and region of search, characteristic element constitutes the characteristic information of describing said static image target.
This method is through search static image target frontier point; Obtain the area information of static image target; Through being worth the factor assessment method, ask for the characteristic element in the frontier point neighborhood, the characteristic information that constitutes static image target by one group of discrete characteristic element and area information.Ask for the characteristic information of static image target, have simply, characteristics fast and intuitively.
And this method can further be evaluated through the target signature element to image, and the target signature element that deletion is redundant makes milder border remember through a spot of characteristic element.Further, the present invention improves the memory effect of still image through increasing the background characteristics element.
Method of the present invention is in the technical field of static image target identification; Solved that prior art is selected the image object characteristic information and acquiring method does not have the problem of versatility, solved prior art and ask for the slow problem that characteristic information produces through the complicated calculations process.Method of the present invention only is used for the memory process to static image target, and need not ask for characteristic information in the identifying of static image target, only needs memory based information, searches for fast, matees in the region of search and can realize Target Recognition.This remarkable result is with the condition of letter property.
The present invention also provides a kind of extraction element of static image target characteristic information, and this device comprises:
The target selection unit is used to select the reference rectangular area of said static image target;
The frontier point acquiring unit is used for beginning with reference to the border, rectangular area from said, searches for along level or vertical inside direction by the SI of setting, and obtains the frontier point of static image target;
The target area parameter calculation unit is used for the RP according to rectangular area, region of search and the characteristic element of said frontier point calculating static image target;
Characteristic element is asked for the unit; Be used to ask for the characteristic element in the said frontier point neighborhood; Said characteristic element comprises the pixel with maximum value factor, and said maximum value factor comprises that 8 neighborhoods are worth factor or 16 neighborhoods are worth the maximum value factor in the factor;
Characteristic information constitutes the unit, is used for the RP of the rectangular area of a stack features element, static image target and region of search, characteristic element is constituted the characteristic information of static image target.
This device obtains the frontier point of static image target through the frontier point acquiring unit, and through target signature element acquiring unit image object frontier point and peripheral neighborhood pixel is worth the factor evaluation, obtains the characteristic element of said image object.Solved that prior art is selected the image object characteristic information and acquiring method does not have the problem of versatility, solved prior art and ask for the slow problem that characteristic information produces through the complicated calculations process.This device can extract discrete image object characteristic element fast, simply and intuitively, and then the more complete image object information of memory.And the present invention can also evaluate the target signature element of image through target signature element filter element, and the characteristic element that deletion is redundant can usually be remembered milder border through more a spot of characteristic element.Further, the present invention obtains the image background characteristic element through background characteristics element acquiring unit, has improved the memory effect of static image target.
Embodiment
In daily life, through vision, mental picture information is our habitual basic act.But, explain the information that you remember, but find, only remembered information seldom.Even like this, do not influence your memory and identification to daily image information.The basic characteristics of thinking in images that Here it is.Simulation thinking in images, the method for distilling that makes up the static image target characteristic information is a core concept of the present invention.Realize memory through a spot of information to the image object principal character.
The embodiment of the invention provides a kind of method for distilling of static image target characteristic information; Frontier point through the search static image target; Obtain the area information of static image target; Through being worth the factor assessment method, ask for the characteristic element in the frontier point neighborhood, the characteristic information that constitutes static image target by one group of discrete characteristic element and area information.Having solved prior art characteristic information acquiring method does not have versatility and asks for slow problem.The embodiment of the invention can be simply, image and obtain the characteristic information of static image target fast.
Specify the characteristics information extraction method embodiment of a kind of static image target of the present invention below in conjunction with accompanying drawing.
As shown in Figure 1, the characteristics information extraction method of a kind of static image target of the embodiment of the invention comprises the steps:
Step 101: select static image target.
In an image, define a static image target through the rectangular area.The rectangular area must intactly comprise the full detail of static image target.If unwanted target information is arranged in the rectangular area, must remove redundant information by wiping figure instrument.This zone is as the reference rectangular area of extracting static image target.
Step 102: the frontier point of search static image target.
Begin from the border, rectangular area, search for along level or vertical inside direction, obtain the frontier point of static image target by the SI of setting.
Step 103: the RP of confirming rectangular area, region of search and the characteristic element of static image target.
In the frontier point coordinate of static image target the search obtain the rectangular area that minimum x coordinate, maximum x coordinate and minimum y coordinate, maximum y coordinate define static image target, be designated as Reb (l, t, r, b).Wherein.The minimum x coordinate of Reb.l=, the minimum y coordinate of Reb.t=, the maximum x coordinate of Reb.r=, the maximum y coordinate of Reb.b=.
Set a rectangular area that comprises whole static image target information, as the region of search of static image target, be designated as Res (l, t, r, b); This zone is used for identifying and confirms the hunting zone.Res can set with the rectangular area, also can setting search scope fa, ask for through following formula:
Res.l=Reb.l-fa,Res.l=Reb.l-fa,Res.l=Reb.l-fa,Res.l=Reb.l-fa。
Ask for the RP of the central point of static image target rectangular area as characteristic element, be designated as C0 (x, y).Wherein, C0.x=(Reb.l+Reb.r)/2, C0.y=(Reb.t+Reb.b)/2.
Step 104: ask for the characteristic element in the frontier point neighborhood; Said characteristic element comprises the pixel with maximum value factor; Said maximum value factor comprises that 8 neighborhoods are worth factor or 16 neighborhoods are worth the maximum value factor in the factor, and as shown in Figure 2, it is following to ask for step:
Step 201: ask for the pixel that has the maximum value factor in the frontier point neighborhood.
The frontier point of the static image target that step 102 is searched is asked for the pixel that has the maximum value factor in the frontier point neighborhood, and as shown in Figure 3, it is following to ask for step:
Step 301:
If the pixel that frontier point takes a step forward along the direction of search is P0;
P0 and P0 periphery 8 neighborhood territory pixels are calculated 8 neighborhoods value factor respectively, obtain maximum value factor J8m and have the pixel that J8m is worth factor;
Step 302:
If J8m<8 are selected near the image object border in the picture element with J8m value factor, near the pixel of the direction of search, the pixel of maximum value factor.
Step 303:
Otherwise, be worth the pixel that factor equals 8 to having, ask 16 neighborhoods to be worth factor respectively, obtain maximum value factor J16m; (step 304) equals 1 if having the pixel number of J16m value factor, gets to have the pixel that J16m is worth factor, as the pixel of maximum value factor; (step 305) otherwise, be worth in the pixel of factor and select having J16m near the image object border, near the pixel of the direction of search, as the pixel of maximum value factor.
Wherein, the acquiring method that 8 neighborhoods and 16 neighborhoods are worth factor comprises: will wait to ask the pixel of value factor to be designated as P0, the inner ring pixel of surrounding P0 has 8, is designated as P1 respectively to P8; The outer ring pixel of surrounding P0 has 16, is designated as P01 respectively to P016;
Be provided with optical parametric decision function Fabs (pi, pj); Judge pi, pj two pixel optics parameter similaritys, if judge similarly, function returns 1, otherwise, return 0;
Utilize the optical parametric decision function respectively, P0 and P1 carried out the optical parametric similarity determination respectively to P8, the sum function rreturn value, obtain 8 neighborhoods value factor J8=∑ Fabs (P0, Pj), (wherein j gets the value of l to 8 in proper order).
Utilize the optical parametric decision function respectively, P0 and P01 carried out the optical parametric similarity determination respectively to P016, the sum function rreturn value, obtain 16 neighborhoods value factor J16=∑ Fabs (P0, Pj), (wherein j gets 01 to 016 value in proper order).
Wherein, optical parametric decision function Fabs (pi, pj) design as follows: be chosen in R, G, B rectangular coordinate color space and carry out the similar judgement of optical parametric.The permission changing value Sa of similarity determination at R, G, B color space is set, and in the color space at pi pixel place be the center a bit, and the cubic space that to set up a length of side be 2 times of Sa is as judgement and the similar space of pi pixel optical parametric.
If abs (pi.r-pj.r)<Sa and abs (pi.g-pj.g)<Sa and abs (pi.b-pj.b)<Sa judge that pi is similar with the pj pixel, return 1; Otherwise, judge that pi and pj pixel are dissimilar, return 0.
In the optical parametric decision function, under the specific (special) requirements situation, the permission changing value of similarity determination color space can be set also, on R, G, B component, be different.
In the optical parametric decision function, also can carry out, for example the cone color space of colourity, color saturation, brightness at other color spaces.By contrast, the calculated amount of optical parametric decision function of the present invention is minimum.
Step 202: will have the coordinate of the pixel of maximum value factor,, will have R, G, the B value of the pixel of maximum value factor, as the optical parametric of characteristic element as the coordinate of characteristic element.
Step 203: through the evaluation to characteristic element, the redundant elements of deleting in the said image object characteristic element is as shown in Figure 4, comprises the steps:
Step 401: with four edges circle is that the search starting point is divided into four groups with characteristic element, and three adjacent characteristic elements of geometric coordinate of select progressively in group are as the computing unit of intermediate features element evaluation;
Step 402: through the optical parametric decision function, characteristic element in the said computing unit is carried out the optical parametric similarity determination,, keep intermediate features element (step 406) if dissimilar;
(step 403) otherwise, judge in the said computing unit that characteristic element whether on same straight line, if not on same straight line, keeps intermediate features element (step 406);
(step 404) otherwise, judge whether there is saltus step information in the said computing unit between characteristic element, if exist, keep intermediate features element (step 406);
(step 405) otherwise, delete said intermediate features element.
Wherein, carry out optical parametric similarity determination and optical parametric decision function Fabs (pi, pj) same as described above.
Whether the characteristic element of judging computing unit on same straight line, can adopt different mathematical methods.For example, method 1: can set apart from permissible error Sc, calculate the distance L c of the intermediate features unit vegetarian refreshments straight line that the characteristic element vegetarian refreshments constitutes to two ends,, otherwise judge not on same straight line if Lc<Sc judges the characteristic element of computing unit on same straight line.Method 2: can set an angle permissible error Sd, calculate the absolute value Ld of intermediate features unit vegetarian refreshments,, otherwise judge not on same straight line if Ld>Sd judges the characteristic element of computing unit on same straight line to the angle of two ends characteristic element vegetarian refreshments.
Judge in the said computing unit; Whether having saltus step information between characteristic element is that whether the optical parametric of the pixel to the straight line of two ends characteristic element vegetarian refreshments is similar with intermediate features unit vegetarian refreshments respectively through judging intermediate features unit vegetarian refreshments; If it is all similar; Judge no saltus step information, otherwise judging there is saltus step information.The method of optical parametric similarity determination is same as described above.
Step 204: as shown in Figure 5, select target is distinguished the level characteristic element in said image object characteristic element, asks for the background characteristics element corresponding with target area graded features element.
Step 501: with four edges circle is that the search starting point is divided into four groups with characteristic element, and the group order of pressing left, up, right, down sorts to characteristic element, the group of each characteristic element of mark, and the number of acquisition characteristic element;
Step 502: target setting is distinguished the selection percentage 1:Nb of level characteristic element, selects the above-mentioned characteristic element that the back sequence number is divided exactly by Nb of arranging again, as target area graded features element;
Step 503: to each target area graded features element, by group, the opposite direction along the ferret out border, search obtains the frontier point of background;
Step 504: ask for the pixel that has the maximum value factor in the said background border vertex neighborhood;
Step 505: with said coordinate with pixel of maximum value factor, the coordinate of characteristic element is asked said R, G, B value with pixel of maximum value factor, as a setting the optical parametric of characteristic element as a setting.
Step 205: with the coordinate conversion of said characteristic element for being the relative coordinate of initial point, as the geometric parameter of characteristic element with the characteristic element RP.The flag information of characteristic element is set: 0: the general features element; 1: target area graded features element; 2: the background characteristics element.
Step 105: the RP by the rectangular area of a stack features element, static image target and region of search, characteristic element constitutes the characteristic information of describing said static image target.
One stack features element comprises the background characteristics element, has described the characteristic of shape and elephant on the static image target border.The array sequence number of characteristic element indicates its serial number, and it is general features element or target area graded features element or background characteristics element that the flag information of characteristic element indicates this element.
The rectangular area of static image target indicates the position of static image target in image; The region of search indicates in identifying, the zone of search static image target.
The RP of characteristic element indicates the center of this static image target and the relativity shift of a reference point.This example is reference point with the true origin.In particular cases, can with the center of other static image target reference point.
Constitute the characteristic information of describing said static image target by above-mentioned information.
According to the method for present embodiment, extract the characteristic information of static image target.Fig. 6 is the synoptic diagram of a stack features element wherein." National Industrial and Commercial Bank of China " several words are static image targets of choosing among the figure, and white point is the clarification of objective element, and the stain of periphery is the background characteristics element.
As shown in Figure 7, the structural drawing for the characteristic information extraction element of static image target in the embodiment of the invention specifically comprises:
Target selection unit 71 is used to select the reference rectangular area of said static image target;
Frontier point acquiring unit 72 is used for beginning with reference to the border, rectangular area from said, searches for along level or vertical inside direction by the SI of setting, and obtains the frontier point of static image target;
Target area parameter calculation unit 73 is used for the RP according to rectangular area, region of search and the characteristic element of said frontier point calculating static image target;
Characteristic element is asked for unit 74; Be used to ask for the characteristic element in the said frontier point neighborhood; Said characteristic element comprises the pixel with maximum value factor, and said maximum value factor comprises that 8 neighborhoods are worth factor or 16 neighborhoods are worth the maximum value factor in the factor;
Characteristic information constitutes unit 75, is used for the RP of the rectangular area of a stack features element, static image target and region of search, characteristic element is constituted the characteristic information of static image target.
Wherein, characteristic element is asked for the unit and is further comprised:
Be worth the factor evaluation unit, be used to ask for the pixel that has the maximum value factor in the said frontier point neighborhood;
The parameter setting unit is used for the coordinate that has the pixel of maximum value factor with said, as the coordinate of characteristic element, with said R, G, B value with pixel of maximum value factor, as the optical parametric of characteristic element;
Target signature element filter element is used for deleting the redundant elements in the said static image target characteristic element according to the evaluation to characteristic element;
Background characteristics element acquiring unit is used for distinguishing the level characteristic element at said static image target characteristic element select target, and asks for and the corresponding background characteristics element of said target area graded features element;
Coordinate transformation unit is used for the coordinate conversion of said characteristic element for being the relative coordinate of initial point with the characteristic element RP, as the geometric parameter of characteristic element, and the flag information of characteristic element is set.
In addition, said device can further include:
Characteristic element is preserved unit 746, is used to preserve the characteristic information of said static image target.
The extraction element of a kind of static image target characteristic information of the embodiment of the invention; Obtain the frontier point of image object through the frontier point acquiring unit; And image object frontier point and peripheral neighborhood pixel are worth the factor evaluation through target signature element acquiring unit; Obtain the image object characteristic element,, realize static image target simply and is fast remembered through discrete target signature element.And the embodiment of the invention can be evaluated the target signature element of image through target signature element filter element, and further the redundant target signature element of deletion can be remembered the milder part in border with target signature element still less.Characteristic element through to image object increases the background characteristics element, can improve the accuracy that static image target is discerned, and improves the antijamming capability of static image target identification.
The above; Be merely embodiment of the present invention, but protection scope of the present invention is not limited thereto, any technician who is familiar with the present technique field is in the technical scope that the present invention discloses; The variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.