CN105335962A - Tobacco field acquisition image segmentation method - Google Patents

Abstract

The invention provides a tobacco field acquisition image segmentation method. According to the method, when that illumination in a scene has a relatively great change is detected, the influence of illumination change on video segmentation is inhibited by utilizing a homomorphic filtering algorithm, and then video segmentation is performed by adopting a color difference histogram algorithm, so that it is ensured that background update and foreground target segmentation still can be steadily performed in a video scene with the relatively great illumination change.

Description

A kind of vega gathers image partition method

Technical field

The present invention relates to image processing field, more specifically, relate to a kind of vega and gather image partition method.

Background technology

To gathering in process that the image of vega processes, how processing the impact that illumination variation brings is an important problem.Although proposed a lot of algorithm for the situation containing illumination variation in scene, but the arbitrariness of illumination variation and rapidity still make these detection algorithms there is higher false drop rate.In recent years, propose a kind of new study mechanism and improve the adaptive faculty of Gaussian Mixture algorithm to background, and utilize based on the prospect false drop rate under the heuristic secondary rate control method minimizing illumination variation of frame difference, also useful 3 threshold values belong to pixel the new method that prospect or background classify, and solve different illumination conditions to a certain extent to the impact of Video segmentation; And the local 3 meta schema operator proposing Scale invariant solves the impact etc. of illumination variation. find that color histogram of difference Video Segmentation that these methods propose has Target Segmentation steadily and surely and the feature such as fast operation compared with above-mentioned algorithm by simulation comparison.Because these algorithms utilize image and its background image color difference to set up color histogram of difference, then segmentation threshold is set, finally judge that each pixel of image belongs to background or foreground target, thus when in scene, illumination variation amplitude is larger, the image inevitably illumination variation caused and the Variant statistical of its background image respective pixel are in color histogram of difference, affect the setting of segmentation threshold, thus cause Iamge Segmentation was lost efficacy, if upgraded unsuccessfully its background image. context update continues to lose efficacy, and will affect subsequent video segmentation effect.

Summary of the invention

The invention provides a kind of vega gather image partition method, the method can suppress illumination variation on the impact of Iamge Segmentation robustness, can be sane carry out image background renewal and foreground object segmentation.

In order to reach above-mentioned technique effect, technical scheme of the present invention is as follows:

A kind of vega gathers image partition method, comprises the following steps:

S1: the vega number of image frames Nf of statistic sampling, adopt the background extraction algorithm based on probability to extract background image f _b(x, y);

S2: extract the i-th two field picture f respectively _ithe redness of (x, y), green and blue valve Ri (x, y), Gi (x, y) and Bi (x, y), i=0,1, Nf;

S3: by the background image f of rgb color space _b(x, y) and the i-th two field picture f _i(x, y) is transformed into HSV color space, and the luminance component of the i-th two field picture and background image thereof is respectively fv _i(x, y) and fv _b(x, y), judges whether the i-th two field picture and background image illumination thereof change, is just split by the i-th two field picture if unchanged, just judges whether the i-th two field picture exists illumination further dim if change;

S4: if it is dim just respectively to fv to there is illumination _i(x, y) and fv _b(x, y) carries out illumination compensation, and by the fv after illumination compensation _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f ' in conjunction with tone, saturation degree component again _i(x, y) and f ' _b(x, y); If it is dim just direct to fv to there is not illumination _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f in conjunction with tone, saturation degree component again " _i(x, y) and f " _b(x, y);

S5: to f ' _i(x, y), f ' _b(x, y) or f " _i(x, y), f " _bthe process that (x, y) repeats step S3-S5 carries out context update and segmentation.

Further, judge in described step S3 the i-th two field picture and the whether vicissitudinous method of background image illumination as follows:

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₁=c ₁× M × N, c ₁for constant coefficient.

Further, the process split the i-th two field picture in described step S3 is as follows:

S31: utilize f _i(x, y) and f _b(x, y) calculates color difference histogram value respectively

ΔR(x,y)＝R _i(x,y)-R _B(x,y)

ΔG(x,y)＝G _i(x,y)-G _B(x,y)x＝1,2,…,M；y＝1,2…,N；

ΔB(x,y)＝B _i(x,y)-B _B(x,y)

S32: to the smoothing process of color difference histogram value:

S33: computed segmentation left and right threshold value SR _i, SL _i:

S34: segmentation image:

S35: if QB (x, y) is 0 just by image f _iin (x, y), the Pixel Information of correspondence position is updated to its background image fv _bin (x, y):

If QB (x, y) is not 0, just removes prospect shade and corresponding QB (x, y) is set to 0;

S36: the hole adopting constraint length algorithm and morphological method to fill up foreground target to QB (x, y) is removed isolated pixel and minimum target and obtained final Iamge Segmentation.

Further, judge in described step S3 that the method whether the i-th two field picture exists illumination dim is as follows:

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₂=c ₂× M × N, c ₂for constant coefficient.

Further, the process of carrying out illumination suppression in described step S4 is as follows:

Further, the process of carrying out illumination compensation in described step S4 is as follows:

Compared with prior art, the beneficial effect of technical solution of the present invention is:

The inventive method is when detecting in scene that illumination has larger change, Homomorphic Filtering Algorithm is utilized to suppress illumination variation on the impact of Video segmentation, then color difference histogramming algorithm is adopted to carry out Video segmentation, thus ensure containing larger illumination variation video scene in, still can be sane carry out context update and foreground object segmentation.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the inventive method.

Embodiment

Accompanying drawing, only for exemplary illustration, can not be interpreted as the restriction to this patent;

In order to better the present embodiment is described, some parts of accompanying drawing have omission, zoom in or out, and do not represent the size of actual product;

To those skilled in the art, in accompanying drawing, some known features and explanation thereof may be omitted is understandable.

Below in conjunction with drawings and Examples, technical scheme of the present invention is described further.

Embodiment 1

A kind of vega gathers image partition method, comprises the following steps:

S1: the vega number of image frames Nf of statistic sampling, adopt the background extraction algorithm based on probability to extract background image f _b(x, y);

S2: extract the i-th two field picture f respectively _ithe redness of (x, y), green and blue valve Ri (x, y), Gi (x, y) and Bi (x, y), i=0,1, Nf;

S3: by the background image f of rgb color space _b(x, y) and the i-th two field picture f _i(x, y) is transformed into HSV color space, and the luminance component of the i-th two field picture and background image thereof is respectively fv _i(x, y) and fv _b(x, y), judges whether the i-th two field picture and background image illumination thereof change, is just split by the i-th two field picture if unchanged, just judges whether the i-th two field picture exists illumination further dim if change;

S4: if it is dim just respectively to fv to there is illumination _i(x, y) and fv _b(x, y) carries out illumination compensation, and by the fv after illumination compensation _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f ' in conjunction with tone, saturation degree component again _i(x, y) and f ' _b(x, y); If it is dim just direct to fv to there is not illumination _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f in conjunction with tone, saturation degree component again " _i(x, y) and f " _b(x, y);

S5: to f ' _i(x, y), f ' _b(x, y) or f " _i(x, y), f " _bthe process that (x, y) repeats step S3-S5 carries out context update and segmentation.

Judge in step S3 the i-th two field picture and the whether vicissitudinous method of background image illumination as follows:

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₁=c ₁× M × N, c ₁for constant coefficient.

The process split the i-th two field picture in step S3 is as follows:

S31: utilize f _i(x, y) and f _b(x, y) calculates color difference histogram value respectively

ΔR(x,y)＝R _i(x,y)-R _B(x,y)

ΔG(x,y)＝G _i(x,y)-G _B(x,y)x＝1,2,…,M；y＝1,2…,N；

ΔB(x,y)＝B _i(x,y)-B _B(x,y)

S32: to the smoothing process of color difference histogram value:

S33: computed segmentation left and right threshold value SR _i, SL _i:

S34: segmentation image:

S35: if QB (x, y) is 0 just by image f _iin (x, y), the Pixel Information of correspondence position is updated to its background image fv _bin (x, y):

If QB (x, y) is not 0, just removes prospect shade and corresponding QB (x, y) is set to 0;

S36: the hole adopting constraint length algorithm and morphological method to fill up foreground target to QB (x, y) is removed isolated pixel and minimum target and obtained final Iamge Segmentation.

Judge in step S3 that the method whether the i-th two field picture exists illumination dim is as follows:

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₂=c ₂× M × N, c ₂for constant coefficient.

The process of carrying out illumination suppression in step S4 is as follows:

Further, the process of carrying out illumination compensation in described step S4 is as follows:

The inventive method is when detecting in scene that illumination has larger change, Homomorphic Filtering Algorithm is utilized to suppress illumination variation on the impact of Video segmentation, then adopt color difference histogramming algorithm to carry out Video segmentation. thus ensure containing in the video scene of larger illumination variation, still can be sane carry out context update and foreground object segmentation.

The corresponding same or analogous parts of same or analogous label;

Describe in accompanying drawing position relationship for only for exemplary illustration, the restriction to this patent can not be interpreted as;

Obviously, the above embodiment of the present invention is only for example of the present invention is clearly described, and is not the restriction to embodiments of the present invention.For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description.Here exhaustive without the need to also giving all embodiments.All any amendments done within the spirit and principles in the present invention, equivalent to replace and improvement etc., within the protection domain that all should be included in the claims in the present invention.

Claims (6)

1. vega gathers an image partition method, it is characterized in that, comprises the following steps:

S1: the vega number of image frames Nf of statistic sampling, adopt the background extraction algorithm based on probability to extract background image f _b(x, y);

S2: extract the i-th two field picture f respectively _ithe redness of (x, y), green and blue valve Ri (x, y), Gi (x, y) and Bi (x, y), i=0,1, Nf;

S3: by the background image f of rgb color space _b(x, y) and the i-th two field picture f _i(x, y) is transformed into HSV color space, and the luminance component of the i-th two field picture and background image thereof is respectively fv _i(x, y) and fv _b(x, y), judges whether the i-th two field picture and background image illumination thereof change, is just split by the i-th two field picture if unchanged, just judges whether the i-th two field picture exists illumination further dim if change;

S4: if it is dim just respectively to fv to there is illumination _i(x, y) and fv _b(x, y) carries out illumination compensation, and by the fv after illumination compensation _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f ' in conjunction with tone, saturation degree component again _i(x, y) and f ' _b(x, y); If it is dim just direct to fv to there is not illumination _i(x, y) and fv _b(x, y) carries out illumination and suppresses to rotate back into rgb space generate f in conjunction with tone, saturation degree component again " _i(x, y) and f " _b(x, y);

S5: to f ' _i(x, y), f ' _b(x, y) or f " _i(x, y), f " _bthe process that (x, y) repeats step S3-S5 carries out context update and segmentation.

2. vega according to claim 1 gathers image partition method, it is characterized in that, judge in described step S3 the i-th two field picture and the whether vicissitudinous method of background image illumination as follows:

$\underset{x=1}{\overset{M}{\Σ}}\underset{y=1}{\overset{N}{\Σ}}|{\mathrm{fv}}_i(x,y)-{\mathrm{fv}}_B(x,y)|>{\mathrm{Th}}_1$

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₁=c ₁× M × N, c ₁for constant coefficient.

3. vega according to claim 1 gathers image partition method, it is characterized in that, the process split the i-th two field picture in described step S3 is as follows:

S31: utilize f _i(x, y) and f _b(x, y) calculates color difference histogram value respectively

ΔR(x,y)＝R _i(x,y)-R _B(x,y)

ΔG(x,y)＝G _i(x,y)-G _B(x,y)x＝1,2,…,M；y＝1,2…,N；

ΔB(x,y)＝B _i(x,y)-B _B(x,y)

S32: to the smoothing process of color difference histogram value:

${\{{\mathrm{HS}}_j^c=\frac{\underset{k=-2}{\overset{2}{\Σ}}{\mathrm{Hist}}_i^c(j+k)}{5}\}}_{c=R,G,B}^{j=-253,-252,...,253};$

S33: computed segmentation left and right threshold value SR _i, SL _i:

$L_t^c=\underset{k\∈(-252,{\mathrm{\μ}}_t^c-1)}{max}\left(ind\right)$

$R_t^c=\underset{k\∈({\mathrm{\μ}}_t^c+1,252)}{min}\left(ind\right)$

$ind\stackrel{\mathrm{\Δ}}{=}\left\{k\right|{\mathrm{HS}}_t^c\left(k\right)\≤{\mathrm{HS}}_t^c(k-1)and{\mathrm{HS}}_t^c\left(k\right)\≤{\mathrm{HS}}_t^c(k+1)\};$

${\mathrm{SR}}_i=R_i^R+R_i^G+R_i^B$

${\mathrm{SL}}_i=L_i^R+L_i^G+L_i^B$

S34: segmentation image:

$QB(x,y)=\left\{\begin{array}{c}0,if{D}_{FB}(x,y)\≤S{R}_{i}or{D}_{FB}(x,y)\\ 1,otherwise\end{array}\right.$

D _FB(x,y)＝|ΔR(x,y)|+|ΔG(x,y)|+|ΔB(x,y)|

S35: if QB (x, y) is 0 just by image f _iin (x, y), the Pixel Information of correspondence position is updated to its background image fv _bin (x, y):

$R_B(x,y)=\frac{R_B(x,y)\×7+R_S(x,y)}{8}$

$G_B(x,y)=\frac{G_B(x,y)\×7+G_S(x,y)}{8}$

$B_B(x,y)=\frac{B_B(x,y)\×7+B_S(x,y)}{8}$

If QB (x, y) is not 0, just removes prospect shade and corresponding QB (x, y) is set to 0;

S36: the hole adopting constraint length algorithm and morphological method to fill up foreground target to QB (x, y) is removed isolated pixel and minimum target and obtained final Iamge Segmentation.

4. vega according to claim 1 gathers image partition method, it is characterized in that, judges that the method whether the i-th two field picture exists illumination dim is as follows in described step S3:

$\underset{x=1}{\overset{M}{\&Sigma;}}\underset{y=1}{\overset{N}{\&Sigma;}}{\mathrm{fv}}_i(x,y)<{\mathrm{Th}}_2$

In formula, x and y represents the pixel coordinate that size is a two field picture of M × N respectively, Th ₂=c ₂× M × N, c ₂for constant coefficient.

5. vega according to claim 1 gathers image partition method, and it is characterized in that, the process of carrying out illumination suppression in described step S4 is as follows:

$\begin{array}{c}{H}_1(u,v)=({\mathrm{\&gamma;}}_H-{\mathrm{\&gamma;}}_L)\&lsqb;1-\exp \left({-c}_0{\left(\frac{D(u,v)}{D_0}\right)}^2\right)\&rsqb;+{\mathrm{\&gamma;}}_L\\ D(u,v)={\&lsqb;{(u-\frac{M}{2})}^2+{(v-\frac{N}{2})}^2\&rsqb;}^{\frac{1}{2}}\end{array}.$

6. vega according to claim 1 gathers image partition method, and it is characterized in that, the process of carrying out illumination compensation in described step S4 is as follows:

$\begin{array}{c}{H}_1(u,v)=({\mathrm{\&gamma;}}_H-{\mathrm{\&gamma;}}_L)\&lsqb;1-\exp \left(-c_0{\left(\frac{D(u,v)}{{D_0}^3}\right)}^2\right)\&rsqb;+{\mathrm{\&gamma;}}_L\\ D(u,v)={\&lsqb;{(u-\frac{M}{2})}^2+{(v-\frac{N}{2})}^2\&rsqb;}^{\frac{1}{2}}\end{array}.$