CN103729834B - The self adaptation joining method of a kind of X ray image and splicing system thereof - Google Patents

Abstract

The invention discloses the self adaptation joining method of a kind of X ray image, specifically include following steps: step 1, obtain the transverse view image-tape of continuous two width X ray images；Step 2, calculate the variance of two transverse view image-tapes respectively and determine template image and mobile image；Step 3, registration template image and mobile image；Image is merged by step 4, employing overlay region non-linear fusion method, obtains spliced new images.The present invention is by comparing the variance of the gray value of the transverse view image-tape chosen respectively from the overlay region part of two width X ray image figures, adaptively selected registrate as template and another width striograph than more visible and that information contained is more transverse view image-tape, and then use the non-linear excessive method in overlapping region to carry out merging by the striograph of input and realize full size splicing.

Description

The self adaptation joining method of a kind of X ray image and splicing system thereof

Technical field

The invention belongs to digital image search technical field, be specifically related to the self adaptation joining method of a kind of X ray image And splicing system.The present invention is for carrying out automatic Mosaic to the X ray image of the zones of different of X ray image equipment collection.

Background technology

What the arriving of digital times made X ray image is digitized into as trend, and X ray image is very easy to doctor The raw focus inspection to patient and location, prejudge the state of an illness and prognosis serves effect greatly.At present, at digital picture Adjustment method such as image enhaucament, rim detection, contours extract, image registration splicing, picture coding etc. have been widely used, to doctor's Work decreases misdiagnosis rate while bringing great convenience.Existing X ray image equipment (such as CR, DR etc.) once shoots Picture size limited (about 17 × 17 inches), obtain full spinal column, full lower limb or whole body complete image at least need shooting Two width have the image of certain overlapping region.During shooting image, in order to not lose informational influence working doctor, generally regulation two width is even The height of the overlay region of continuous image more than 250 pixels, in order to more comprehensively, get more information about patient's focus and around position thereof Situation, needs to carry out these images the seamless spliced image obtaining super large visual angle.

Traditional digital picture stitching algorithm mainly includes image registration and two steps of image co-registration.First, image is joined Quasi-algorithm is broadly divided into the registration Algorithm of feature based and registration Algorithm based on half-tone information, due to the feature of X ray image Point is the most less, and the method for registering images of feature based often precision does not reaches the required precision of medical figure registration, therefore, and mesh The study hotspot of front medical figure registration is concentrated mainly on registration Algorithm based on half-tone information, and based on normalization in this algorithm The registration Algorithm of mutual information is widely used because having preferable robustness and accuracy, but, traditional normalizing Change mutual information method image registration algorithm and choose the part of wherein piece image overlapping region regularly as mould when registrating image Plate image registrates with another piece image, and two width images are different in the image information contained by overlapping region, so Registration result also has a greater change with the difference of stencil-chosen to cause final result.Secondly, common image interfusion method Having mean value method and overlay region linearly excessive method, the most conventional method is overlay region linearly excessive method, and it uses linear Conversion weight carry out fusion image so that transition region seems milder, but, the method cannot fully eliminate two subject to registration The impact on image registration accuracy such as image border part overexposure, image border part excessive deformation.

Summary of the invention

For defect or deficiency present in above-mentioned prior art, it is an object of the invention to, it is provided that one is applied to X and penetrates The self adaptation joining method of line image and splicing system thereof, the present invention is by comparing respectively from the overlap of two width X ray image figures The variance of the gray value of the transverse view image-tape that district's part is chosen, adaptively selected than more visible and that information contained is more transverse view Image-tape registrates as template and another width striograph, and then uses the non-linear excessive method in overlapping region by the striograph of input Carry out merging and realize full size splicing.

In order to achieve the above object, the present invention adopts the following technical scheme that and is solved:

The self adaptation joining method of a kind of X ray image, specifically includes following steps:

Step 1, obtain the transverse view image-tape of continuous two width X ray images；

Step 2, calculate the variance of two transverse view image-tapes respectively and determine template image and mobile image；

Step 3, registration template image and mobile image；

Image is merged by step 4, employing overlay region non-linear fusion method, obtains spliced new images.

Further, the concrete operations of described step 1 are as follows:

Order two X ray image images to be spliced of reading: the first width striograph I₁With the second width striograph I₂；? One width striograph I₁On with the second width striograph I₂Overlapping region intercept transverse view image-tape I_u, at the second width striograph I₂On With the first width striograph I₁Overlapping region intercept transverse view image-tape I_d, it is ensured that transverse view image-tape I_uWith transverse view image-tape I_d's Width is equal and is equal to the width of X ray image image to be spliced, ensures transverse view image-tape I simultaneously_uWith transverse view image-tape I_d's The most equal and no more than 250 pixels.

Further, the concrete operations of described step 2 are as follows:

1) transverse view image-tape I is calculated respectively according to formula 6, formula 7_uVarianceWith transverse view image-tape I_dVariance:

$\left\{\begin{array}{c}{\mathrm{\μ}}_u=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_u}{f}_u(x,y)\\ {\mathrm{\σ}}_u^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_u}{[f_u(x,y)-{\mathrm{\μ}}_u]}^2\end{array}\right.---\left(6\right)$

$\left\{\begin{array}{c}{\mathrm{\μ}}_d=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_d}{f}_d(x,y)\\ {\mathrm{\σ}}_d^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_d}{[f_d(x,y)-{\mathrm{\μ}}_d]}^2\end{array}\right.---\left(7\right)$

Wherein, f_u(x y) is transverse view image-tape I_uAt coordinate position (x, y) gray value at place；f_d(x y) is landscape images Band I_dAt coordinate position (x, y) gray value at place；W_tFor transverse view image-tape I_uWith transverse view image-tape I_dWidth；H_tFor landscape images Band I_uWith transverse view image-tape I_dHeight；

2) ifIt is more thanChoose striograph I₁For template image, striograph I₂For mobile image；Otherwise, choose laterally Picture strip I_dFor template image, striograph I₁For mobile image.

Further, the concrete operations of described step 3 are as follows:

Template image is made to translate pixel-by-pixel on mobile image, often one pixel of translation, then take and mould on mobile image The image section that plate image is corresponding, and calculate the normalized mutual information between mobile image and template image；Until mobile image Upper all pixels are traversed；Relatively obtaining the position at template image place during normalized mutual information maximum, record is when this position Template image translational movement Δ away from initial traverse position in vertical direction, obtains striograph I by formula 8₁With striograph I₂Weight The height D in folded district:

D=Δ+H_t(8)

In formula, H_tFor transverse view image-tape I_uWith transverse view image-tape I_dHeight；

Further, the concrete operations of described step 4 are as follows:

If striograph I₁Height be H₁, striograph I₂Height be H₂, creating a fabric width degree is W_d, height be H₁+H₂-D New images, for i.e. height above overlay region from 1 to H₁The part of-D, by striograph I₁In corresponding part copy to newly scheme The correspondence position of picture；For i.e. height below overlay region from H₁+ 1 arrives H₁+H₂-D part, by striograph I₂In corresponding part multiple System is to the correspondence position of new images；

It is that (x, some y) find this point corresponding to striograph I for internal coordinate position, overlay region₁Pixel value f₁(x,y) And this point is corresponding to striograph I₂Pixel value f₂(x y), utilizes formula 9 to be calculated the internal coordinate position, overlay region of new images For (x, (x y), finally gives new images to the pixel value f of some y).

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{(\frac{y}{D}-\frac{1}{2})}^{\frac{1}{3}}+\frac{1}{2}\\ f(x,y)=(1-\mathrm{\ω})\×f_1(x,y)+\mathrm{\ω}\×f_2(x,y)\end{array}\right.---\left(9\right).$

Apply the splicing system of the self adaptation joining method of above-mentioned X ray image, specifically include to be sequentially connected as follows and connect Module:

Image input and transverse view image-tape acquisition module: for obtaining the transverse view image-tape of continuous two width X ray images；

Template image and mobile image determine module: for calculating the variance of two transverse view image-tapes respectively and determining template Image and mobile image；

Template image and mobile image registration module: be used for registrating template image and mobile image；

Image co-registration module: be used for using overlay region non-linear fusion method that image is merged, obtain spliced newly Image.

Further, the input of described image and transverse view image-tape acquisition module are specifically for realizing following functions:

Order two X ray image images to be spliced of reading: the first width striograph I₁With the second width striograph I₂；? One width striograph I₁On with the second width striograph I₂Overlapping region intercept transverse view image-tape I_u, at the second width striograph I₂On With the first width striograph I₁Overlapping region intercept transverse view image-tape I_d, it is ensured that transverse view image-tape I_uWith transverse view image-tape I_d's Width is equal and is equal to the width of X ray image image to be spliced, ensures transverse view image-tape I simultaneously_uWith transverse view image-tape I_d's The most equal and no more than 250 pixels.

Further, described template image and mobile image determine that module is specifically for realizing following functions:

1) transverse view image-tape I is calculated respectively according to formula 6, formula 7_uVarianceWith transverse view image-tape I_dVariance

$\left\{\begin{array}{c}{\mathrm{\μ}}_u=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_u}{f}_u(x,y)\\ {\mathrm{\σ}}_u^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_u}{[f_u(x,y)-{\mathrm{\μ}}_u]}^2\end{array}\right.---\left(6\right)$

$\left\{\begin{array}{c}{\mathrm{\μ}}_d=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_d}{f}_d(x,y)\\ {\mathrm{\σ}}_d^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_d}{[f_d(x,y)-{\mathrm{\μ}}_d]}^2\end{array}\right.---\left(7\right)$

Wherein, f_u(x y) is transverse view image-tape I_uAt coordinate position (x, y) gray value at place；f_d(x y) is landscape images Band I_dAt coordinate position (x, y) gray value at place；W_tFor transverse view image-tape I_uWith transverse view image-tape I_dWidth；H_tFor landscape images Band I_uWith transverse view image-tape I_dHeight；

2) ifIt is more thanChoose striograph I₁For template image, striograph I₂For mobile image；Otherwise, choose laterally Picture strip I_dFor template image, striograph I₁For mobile image.

Further, described template image and mobile image registration module are specifically for realizing following functions:

Template image is made to translate pixel-by-pixel on mobile image, often one pixel of translation, then take and mould on mobile image The image section that plate image is corresponding, and calculate the normalized mutual information between mobile image and template image；Until mobile image Upper all pixels are traversed；Relatively obtaining the position at template image place during normalized mutual information maximum, record is when this position Template image translational movement Δ away from initial traverse position in vertical direction, obtains striograph I by formula 8₁With striograph I₂Weight The height D in folded district:

D=Δ+H_t(8)

In formula, H_tFor transverse view image-tape I_uWith transverse view image-tape I_dHeight；

Further, described image co-registration module is specifically for realizing following functions:

If striograph I₁Height be H₁, striograph I₂Height be H₂, creating a fabric width degree is W_d, height be H₁+H₂-D New images, for i.e. height above overlay region from 1 to H₁The part of-D, by striograph I₁In corresponding part copy to newly scheme The correspondence position of picture；For i.e. height below overlay region from H₁+ 1 arrives H₁+H₂-D part, by striograph I₂In corresponding part multiple System is to the correspondence position of new images；

It is that (x, some y) find this point corresponding to striograph I for internal coordinate position, overlay region₁Pixel value f₁(x,y) And this point is corresponding to striograph I₂Pixel value f₂(x y), utilizes formula 9 to be calculated the internal coordinate position, overlay region of new images For (x, (x y), finally gives new images to the pixel value f of some y)；

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{(\frac{y}{D}-\frac{1}{2})}^{\frac{1}{3}}+\frac{1}{2}\\ f(x,y)=(1-\mathrm{\ω})\×f_1(x,y)+\mathrm{\ω}\×f_2(x,y)\end{array}\right.---\left(9\right).$

Compared with traditional X-ray joining method and system, advantages of the present invention is as follows:

First, the present invention calculates two width image section overlapping region information contained respectively, selects more visible and information contained More parts of images overlay region registrates with another width imaged image as template, it is possible to reach more accurate adaptively Registration result, it is to avoid stencil-chosen is depended on unduly by the final result of traditional normalized mutual information method image registration algorithm Property.

Second, the present invention, during carrying out image co-registration, selects the non-linear transition in overlay region fusion method, by overlap Region carry out merging time weight change be nonlinear so that the image border part weight shared when image co-registration is more Few, so that overlapping region is more smooth, it is possible to preferably suppression region, image border overexposure is shaken during having been taken by Deng impact, it is to avoid overlay region linearly excessively method cannot fully eliminate image border part overexposure and excessive deformation to registration accuracy Impact.

Accompanying drawing explanation

Fig. 1 is the flow chart of the self adaptation stitching algorithm of the X ray image of the present invention.

Fig. 2 is the structural representation of the self adaptation splicing system of the X ray image of the present invention.

Fig. 3 is the effect contrast figure of embodiments of the invention 1.Wherein, (a) is from top to bottom followed successively by three width to be spliced Continuous print X ray image image；B X ray image image that () obtains after splicing for using the method for the present invention.

Fig. 4 is the step 3 calculated stack height D of embodiments of the invention 1.In figure, overlapping region is all straight by level Line marks.

Fig. 5 is in the step 3 of embodiments of the invention 1, template image and the schematic diagram of mobile process of image registration.Its In, (a) and (b), (c) are respectively and are moved the most pixel-by-pixel on template image by movement image and position normalization mutual trust The process of the position that breath is maximum.All including three horizontal lines in (a) and (b), (c) in the figure on the left side, these three horizontal lines are by up to At the beginning of the lower final position of template image matching process of expression respectively, the position of normalized mutual information maximum, template image coupling Beginning position.

Fig. 6 is directly to carry out image mosaic and do not carry out image co-registration and carried out overlay region according to the step 4 of the present invention Contrast on effect before and after non-linear fusion.Wherein, (a) represents the effect of two width consecutive image direct splicing；B () is according to this Bright step 4 has carried out the effect before and after the non-linear fusion of overlay region.

Fig. 7 is the body side X ray image image adaptive splicing effect figure of embodiments of the invention 2.Wherein, (a) In be from top to bottom followed successively by three width continuous print X ray image images to be spliced；B () splices for using the method for the present invention The X ray image image obtained afterwards.

Detailed description of the invention

First, the relevant technical terms related in the present invention is described below:

1, the variance of gradation of image: i.e. weigh a kind of value of image all pixels gray value fluctuation size, variance is more Greatly, showing that the intensity value ranges of this image is the widest, gray-value variation is the biggest, i.e. information contained by this image is the most.? When registrating X ray image figure, alternative template is the part weight of upper graph picture in the two width imaged images loaded Folded district or the district that partly overlaps of lower edge image.Which width striograph is enough clear to select which overlay region to depend on, information contained More, i.e. the variance of image intensity value is the biggest.When registrating with preferable edge, it is possible to reach more preferable registration accuracy.

(x is y) gray level image I at coordinate position for (W is the width of gray level image for x, gray value y), and H be grey to note f The height of degree image, μ is the average of gray-scale map, then the variances sigma of the gray scale of gray level image I²Computing formula as follows:

$\left\{\begin{array}{c}\mathrm{\μ}=\frac{1}{W\×H}{\mathrm{\Σ}}_{(x,y)\∈I}f(x,y)\\ {\mathrm{\σ}}^2=\frac{1}{W\×H-1}{\mathrm{\Σ}}_{(x,y)\∈I}{[f(x,y)-\mathrm{\μ}]}^2\end{array}\right.---\left(1\right)$

2, normalized mutual information (Normalized Mutual Information): for describing the system between two systems Included in meter dependency, or a system in another system information number, it can use entropy (Entropy) to retouch State.What entropy was expressed is complexity or the uncertainty of a system, when the locus of two width images reaches completely the same Time, the information about another piece image that wherein piece image is expressed, namely the mutual information of respective pixel gray scale should be maximum. For discrete digital picture, the mutual information of two width gray-scale maps can represent with normalized joint histogram, for two fabric width degree It is W', is highly an equal amount of image A and B of H', remember M_AI () is that in image A, gray value is the individual of the pixel of i Number, M_BJ () is that in image B, gray value is the number of the pixel of j, M_AB(i is j) that in image A, gray value is that i is simultaneously at image B Middle gray value is the pixel number of j, then the grey level histogram h of image A_AThe grey level histogram h of (i) and image B_B(j) and it Associating grey level histogram be expressed as:

$\left\{\begin{array}{c}{h}_A\left(i\right)=\frac{M_A\left(i\right)}{W^{\′}\×H^{\′}}\\ h_B\left(j\right)=\frac{M_B\left(j\right)}{W^{\′}\×H^{\′}}\\ h_{\mathrm{AB}}(i,j)=\frac{M_{\mathrm{AB}}(i,j)}{W^{\′}\×H^{\′}}\end{array}\right.---\left(2\right)$

Formula 3 is utilized to calculate border entropy H (A) of image A and image B and combination entropy H (A, B) of H (B) and two width images:

$\left\{\begin{array}{c}H\left(A\right)=-\underset{i}{\mathrm{\Σ}}{h}_A\left(i\right)\log \left(h_A\left(i\right)\right)\\ H\left(B\right)=-\underset{j}{\mathrm{\Σ}}{h}_B\left(j\right)\log \left(h_B\left(j\right)\right)\\ H(A,B)=-\underset{i}{\mathrm{\Σ}}\underset{j}{\mathrm{\Σ}}{h}_{\mathrm{AB}}(i,j)\log \left(h_{\mathrm{AB}}(i,j)\right)\end{array}\right.---\left(3\right)$

Formula 4 is utilized to calculate the normalized mutual information between image A and image B:

$I(A,B)=\frac{H\left(A\right)+H\left(B\right)}{H(A,B)}---\left(4\right)$

3, overlay region non-linear fusion: in order to after making fusion, Liang Fu picture registration district is more smooth and being different from of using The image interfusion method of linear weight.Overlay region between needing them after continuous print two width striograph is registrated Merging, traditional image co-registration is when the gray value of the image calculated after merging, it is believed that distance is spliced in image The heart is the most remote, and corresponding weight is the least, but weight is linear transitions, if using nonlinear transition weight, it is possible to more preferably Utilize in original image the half-tone information away from image border part.To two width continuous image figure I₁And I₂Utilize normalization After mutual information method registrates, can get the height D of overlapping region between them, for internal coordinate position, overlay region be (x, some y) find this point corresponding to striograph I₁Pixel value f₁(x, y) and this point is corresponding to striograph I₂Pixel value f₂(x, y), utilizes formula 5 to calculate nonlinear weight ω, obtains internal coordinate position, fusion image overlay region for (x, the pixel of some y) Value f (x, y):

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{(\frac{y}{D}-\frac{1}{2})}^{\frac{1}{3}}+\frac{1}{2}\\ f(x,y)=(1-\mathrm{\ω})\×f_1(x,y)+\mathrm{\ω}\×f_2(x,y)\end{array}\right.---\left(5\right)$

As it is shown in figure 1, the self adaptation joining method of the X ray image of the present invention, specifically include following steps:

Step 1, obtain the transverse view image-tape of continuous two width X ray images.Concrete operations are as follows:

Order two X ray image images to be spliced of reading: the first width striograph I₁With the second width striograph I₂(this Think in bright that the width of continuous print X ray image image to be spliced is equal).At the first width striograph I₁On with the second width shadow As figure I₂Overlapping region intercept transverse view image-tape I_u, at the second width striograph I₂On with the first width striograph I₁Overlay region Territory intercepts transverse view image-tape I_d, it is ensured that transverse view image-tape I_uWith transverse view image-tape I_dWidth equal and be equal to X-ray to be spliced The width of imaged image, ensures transverse view image-tape I simultaneously_uWith transverse view image-tape I_dHeight equal.

Step 2, calculate the variance of two transverse view image-tapes and determine template image respectively.Concrete operations are as follows:

1) transverse view image-tape I is calculated respectively according to formula 6, formula 7_uVarianceWith transverse view image-tape I_dVariance

$\left\{\begin{array}{c}{\mathrm{\μ}}_u=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_u}{f}_u(x,y)\\ {\mathrm{\σ}}_u^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_u}{[f_u(x,y)-{\mathrm{\μ}}_u]}^2\end{array}\right.---\left(6\right)$

$\left\{\begin{array}{c}{\mathrm{\μ}}_d=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_d}{f}_d(x,y)\\ {\mathrm{\σ}}_d^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_d}{[f_d(x,y)-{\mathrm{\μ}}_d]}^2\end{array}\right.---\left(7\right)$

Wherein, f_u(x y) is transverse view image-tape I_uAt coordinate position (x, y) gray value at place；f_d(x y) is landscape images Band I_dAt coordinate position (x, y) gray value at place；W_tFor transverse view image-tape I_uWith transverse view image-tape I_dWidth；H_tFor landscape images Band I_uWith transverse view image-tape I_dHeight；

2) ifIt is more thanChoose striograph I₁For template image, striograph I₂For mobile image；Otherwise, choose laterally Picture strip I_dFor template image, striograph I₁For mobile image；

Step 3, registration template image and mobile image.Concrete operations are as follows:

Template image is made to translate pixel-by-pixel on mobile image, often one pixel of translation, then take and mould on mobile image The image section that plate image is corresponding, and calculate the normalized mutual information between mobile image and template image；Until mobile image Upper all pixels are traversed；Relatively obtaining the position at template image place during normalized mutual information maximum, record is when this position Template image translational movement Δ away from initial traverse position in vertical direction, obtains striograph I by formula 8₁With striograph I₂Weight The height D in folded district:

D=Δ+H_t(8)

In formula, H_tFor transverse view image-tape I_uWith transverse view image-tape I_dHeight；

Image is merged by step 4, employing overlay region non-linear fusion method.Concrete operations are as follows:

If striograph I₁Height be H₁, striograph I₂Height be H₂, creating a fabric width degree is W_d, height be H₁+H₂-D New images, for i.e. height above overlay region from 1 to H₁The part of-D, by striograph I₁In corresponding part copy to newly scheme The correspondence position of picture；For i.e. height below overlay region from H₁+ 1 arrives H₁+H₂-D part, by striograph I₂In corresponding part multiple System is to the correspondence position of new images；

It is that (x, some y) find this point corresponding to striograph I for internal coordinate position, overlay region₁Pixel value f₁(x,y) And this point is corresponding to striograph I₂Pixel value f₂(x y), utilizes formula 9 to be calculated the internal coordinate position, overlay region of new images For (x, (x y), finally gives new images to the pixel value f of some y).

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{(\frac{y}{D}-\frac{1}{2})}^{\frac{1}{3}}+\frac{1}{2}\\ f(x,y)=(1-\mathrm{\ω})\×f_1(x,y)+\mathrm{\ω}\×f_2(x,y)\end{array}\right.---\left(9\right)$

In order to the effectiveness of the method for the present invention is described, inventor has carried out the test of X ray image image mosaic.Use Operating system be Win764, CPU be Intel (R) Core (TM) [email protected].

Embodiment 1:

The present embodiment carries out registration and merges, as shown in Fig. 3 (a) the radioscopic image in continuous print three width human body front.Respectively Choose a transverse view image-tape and the top edge overlay region of the second width X image of the lower limb overlay region of the first secondary radioscopic image A transverse view image-tape, the height of two transverse view image-tapes is equal and is 250 pixels；Then this two picture strips are compared The variance of gray value, chooses the bigger picture strip of variance and compares as template image and another piece image.In this embodiment In, choose the top edge overlay region of the second secondary radioscopic image as template image, choose piece image for mobile image, will The normalized mutual information of template image and the first width imagery exploitation step 3 compares and registrates, and finally uses overlapping region non-thread Property fusion method to registration image merge.Second width and the 3rd width process of image registration ibid, registration result such as Fig. 3 (b) institute Show.By Fig. 3 it will be seen that use step 4 carried out overlapping region merge after, the overexposure at image border to be spliced Situation is enhanced, and obtains the spliced image that the transition shown in Fig. 3 (b) is soft, it is known that the method for the present invention suppresses effectively The impacts such as marginal portion overexposure or shooting shake.

Embodiment 2:

The present embodiment is identical with embodiment, differs only in the present embodiment adaptive to the X ray image image of body side Should splice, before splicing, image is shown in Fig. 7 with the contrast effect of spliced image.

Inventor by carrying out splicing test to 100 groups of continuous print X ray image images, and the result obtained is reliable and stable, Average handling time is 279.6ms, it can be seen that the method for the present invention can not only the most mildly splice X ray image, and Processing speed is fast, and operational efficiency is high.

Claims (2)

1. the self adaptation joining method of an X ray image, it is characterised in that specifically include following steps:

Step 1, obtain the transverse view image-tape of continuous two width X ray images；The concrete operations of described step 1 are as follows:

Order two X ray image images to be spliced of reading: the first width striograph I₁With the second width striograph I₂；At the first width Striograph I₁On with the second width striograph I₂Overlapping region intercept transverse view image-tape I_u, at the second width striograph I₂On with First width striograph I₁Overlapping region intercept transverse view image-tape I_d, it is ensured that transverse view image-tape I_uWith transverse view image-tape I_dWidth Equal and be equal to the width of X ray image image to be spliced, ensure transverse view image-tape I simultaneously_uWith transverse view image-tape I_dHeight Equal；

Step 2, calculate the variance of two transverse view image-tapes respectively and determine template image and mobile image；The tool of described step 2 Gymnastics is made as follows:

1) transverse view image-tape I is calculated respectively according to formula 6, formula 7_uVarianceWith transverse view image-tape I_dVariance

$\left\{\begin{array}{c}{\mathrm{\μ}}_u=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_u}{f}_u(x,y)\\ {\mathrm{\σ}}_u^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_u}{\[f_u(x,y)-{\mathrm{\μ}}_u\]}^2\end{array}\right.---\left(6\right)$

$\left\{\begin{array}{c}{\mathrm{\μ}}_d=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_d}{f}_d(x,y)\\ {\mathrm{\σ}}_d^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_d}{\[f_d(x,y)-{\mathrm{\μ}}_d\]}^2\end{array}\right.---\left(7\right)$

Wherein, f_u(x y) is transverse view image-tape I_uAt coordinate position (x, y) gray value at place；f_d(x y) is transverse view image-tape I_d? Coordinate position (x, y) gray value at place；W_tFor transverse view image-tape I_uWith transverse view image-tape I_dWidth；H_tFor transverse view image-tape I_uWith Transverse view image-tape I_dHeight；

2) ifIt is more thanChoose striograph I₁For template image, striograph I₂For mobile image；Otherwise, landscape images is chosen Band I_dFor template image, striograph I₁For mobile image；

Step 3, registration template image and mobile image；The concrete operations of described step 3 are as follows:

Template image is made to translate pixel-by-pixel on mobile image, often one pixel of translation, then take and Prototype drawing on mobile image As corresponding image section, and calculate the normalized mutual information between mobile image and template image；Until institute on mobile image Pixel is had to be traversed；Relatively obtain the position at template image place during normalized mutual information maximum, record template when this position Image translational movement Δ away from initial traverse position in vertical direction, obtains striograph I by formula 8₁With striograph I₂Overlay region Height D:

D=Δ+H_t (8)

In formula, H_tFor transverse view image-tape I_uWith transverse view image-tape I_dHeight；

Image is merged by step 4, employing overlay region non-linear fusion method, obtains spliced new images；Described step 4 Concrete operations are as follows:

If striograph I₁Height be H₁, striograph I₂Height be H₂, creating a fabric width degree is W_d, height be H₁+H₂The new figure of-D Picture, for i.e. height above overlay region from 1 to H₁The part of-D, by striograph I₁In corresponding part copy to the right of new images Answer position；For i.e. height below overlay region from H₁+ 1 arrives H₁+H₂-D part, by striograph I₂In corresponding part copy to newly The correspondence position of image；

It is that (x, some y) find this point corresponding to striograph I for internal coordinate position, overlay region₁Pixel value f₁(x, y) and This point is corresponding to striograph I₂Pixel value f₂(x, y), utilize formula 9 be calculated the internal coordinate position, overlay region of new images for (x, (x y), finally gives new images to the pixel value f of point y)；

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{\left(\frac{y}{D}-\frac{1}{2}\right)}^{\frac{1}{3}}+\frac{1}{2}\\ f\left(x,y\right)=\left(1-\mathrm{\ω}\right)\×f_1\left(x,y\right)+\mathrm{\ω}\×f_2\left(x,y\right)\end{array}\right.---\left(9\right).$

2. the splicing system of the self adaptation joining method of application X ray image described in claim 1, it is characterised in that concrete Including being sequentially connected the module connect as follows:

Image input and transverse view image-tape acquisition module: for obtaining the transverse view image-tape of continuous two width X ray images；Described figure As input and transverse view image-tape acquisition module are specifically for realizing following functions:

Order two X ray image images to be spliced of reading: the first width striograph I₁With the second width striograph I₂；At the first width Striograph I₁On with the second width striograph I₂Overlapping region intercept transverse view image-tape I_u, at the second width striograph I₂On with First width striograph I₁Overlapping region intercept transverse view image-tape I_d, it is ensured that transverse view image-tape I_uWith transverse view image-tape I_dWidth Equal and be equal to the width of X ray image image to be spliced, ensure transverse view image-tape I simultaneously_uWith transverse view image-tape I_dHeight Equal and no more than 250 pixels；

Template image and mobile image determine module: for calculating the variance of two transverse view image-tapes respectively and determining template image With mobile image；Described template image and mobile image determine that module is specifically for realizing following functions:

1) transverse view image-tape I is calculated respectively according to formula 6, formula 7_uVarianceWith transverse view image-tape I_dVariance

$\left\{\begin{array}{c}{\mathrm{\μ}}_u=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_u}{f}_u(x,y)\\ {\mathrm{\σ}}_u^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_u}{\[f_u(x,y)-{\mathrm{\μ}}_u\]}^2\end{array}\right.---\left(6\right)$

$\left\{\begin{array}{c}{\mathrm{\μ}}_d=\frac{1}{W_t\×H_t}{\mathrm{\Σ}}_{(x,y)\∈I_d}{f}_d(x,y)\\ {\mathrm{\σ}}_d^2=\frac{1}{W_t\×H_t-1}{\mathrm{\Σ}}_{(x,y)\∈I_d}{\[f_d(x,y)-{\mathrm{\μ}}_d\]}^2\end{array}\right.---\left(7\right)$

Wherein, f_u(x y) is transverse view image-tape I_uAt coordinate position (x, y) gray value at place；f_d(x y) is transverse view image-tape I_d? Coordinate position (x, y) gray value at place；W_tFor transverse view image-tape I_uWith transverse view image-tape I_dWidth；H_tFor transverse view image-tape I_uWith Transverse view image-tape I_dHeight；

2) ifIt is more thanChoose striograph I₁For template image, striograph I₂For mobile image；Otherwise, landscape images is chosen Band I_dFor template image, striograph I₁For mobile image；

Template image and mobile image registration module: be used for registrating template image and mobile image；Described template image and movement Image registration module is specifically for realizing following functions:

Template image is made to translate pixel-by-pixel on mobile image, often one pixel of translation, then take and Prototype drawing on mobile image As corresponding image section, and calculate the normalized mutual information between mobile image and template image；Until institute on mobile image Pixel is had to be traversed；Relatively obtain the position at template image place during normalized mutual information maximum, record template when this position Image translational movement Δ away from initial traverse position in vertical direction, obtains striograph I by formula 8₁With striograph I₂Overlay region Height D:

D=Δ+H_t (8)

In formula, H_tFor transverse view image-tape I_uWith transverse view image-tape I_dHeight；

Image co-registration module: be used for using overlay region non-linear fusion method that image is merged, obtain spliced new images； Described image co-registration module is specifically for realizing following functions:

If striograph I₁Height be H₁, striograph I₂Height be H₂, creating a fabric width degree is W_d, height be H₁+H₂The new figure of-D Picture, for i.e. height above overlay region from 1 to H₁The part of-D, by striograph I₁In corresponding part copy to the right of new images Answer position；For i.e. height below overlay region from H₁+ 1 arrives H₁+H₂-D part, by striograph I₂In corresponding part copy to newly The correspondence position of image；

It is that (x, some y) find this point corresponding to striograph I for internal coordinate position, overlay region₁Pixel value f₁(x, y) and This point is corresponding to striograph I₂Pixel value f₂(x, y), utilize formula 9 be calculated the internal coordinate position, overlay region of new images for (x, (x y), finally gives new images to the pixel value f of point y)；

$\left\{\begin{array}{c}\mathrm{\ω}=2^{-\frac{2}{3}}{\left(\frac{y}{D}-\frac{1}{2}\right)}^{\frac{1}{3}}+\frac{1}{2}\\ f\left(x,y\right)=\left(1-\mathrm{\ω}\right)\×f_1\left(x,y\right)+\mathrm{\ω}\×f_2\left(x,y\right)\end{array}\right.---\left(9\right).$