CN103810716B - Move and the image partition method of Renyi entropy based on gray scale - Google Patents

Abstract

The invention discloses and a kind of move and the image partition method of Renyi entropy based on gray scale, including: the video flowing of aircraft collection is extracted image f, and (x, y), based on image f, (x, the luminance component in y) carries out statistics of histogram；Grey level histogram is carried out gray scale move；The grey level histogram of moving obtained after gray scale is moved carries out the binary image segmentation based on Renyi entropy, obtains correctly distinguishing the bianry image of barrier zone and background.The present invention grey level histogram to being counted by original image has carried out gray scale and has moved improvement, then the grey level histogram of moving obtained after gray scale being moved carries out the binary image segmentation based on Renyi entropy, to obtain correctly distinguishing the bianry image of barrier zone and background, the region thus can effectively remove barriers, is suitable to, for finding, the flat site reliable foundation of offer that aircraft security lands.

Description

Move and the image partition method of Renyi entropy based on gray scale

Technical field

The present invention relates to and a kind of move and the image partition method of Renyi entropy based on gray scale, belong to image processing field.

Background technology

When aircraft (unmanned plane) lands in advance, can start to gather video image from 3000 meters of high-altitudes by the monitoring camera installed on it, find flat site in the picture, land for aircraft security.For the image gathered, this image potentially includes the scenery such as waters, building, highway, therefore, for safe landing, must go out in real time, exactly by above-mentioned scene detection, carry out image segmentation in real time, just be provided that out for the flat site that aircraft security lands.

At present, image segmentation can be divided into transform domain and two kinds of dividing methods of time domain.

The dividing method of transform domain is all if any wavelet transformation, Fourier transformation etc., and algorithm is complex, it is difficult to run under the real time environment of DSP.

The dividing method of time domain realizes simply, mainly including edge detection method, thresholding method etc. relatively.

Edge is the feature that image is most basic, and rim detection plays an important role in computer vision, graphical analysis etc. are applied, and is the important step of graphical analysis and identification.Splitting based on the image at edge is a kind of Main Means of graphical analysis and pattern recognition, but edge detection method is suitable for the image that noise is less, less complicated, and the image in the area such as the Gobi desert complex for physical features environment or desert is also inapplicable.

Thresholding method is a kind of traditional, the most frequently used image partition method, it realizes simply, amount of calculation is little, performance is more stable, having become most basic and most widely used a kind of cutting techniques in image segmentation, it is particularly well-suited to target and background and occupies the image of different grey-scale scope.Image threshold be intended to according to gray level, collection of pixels is carried out a division, such division can realize by choosing one or more threshold value from gray level.Typical thresholding method mainly has iterative method, Da-Jin algorithm, Threshold segmentation 0tsu method and Renyi entropy method.It is found that the optimal threshold that solves under complex scene of iterative method and Da-Jin algorithm is less than normal in actual emulation process, the image being partitioned into has more noise, it is impossible to practical requirement.And other thresholding method can not obtain good segmentation effect for the image in the areas such as the complex Gobi desert of physical features environment or desert.

As can be seen here, the image designing the areas such as a kind of Gobi desert that physical features environment is complex or desert carries out effective image segmentation, in order to the technical scheme searching out flat site for aircraft security landing is current urgent problem.

Summary of the invention

It is an object of the invention to provide and a kind of move and the image partition method of Renyi entropy based on gray scale, the method grey level histogram to being counted by original image carries out gray scale and moves, then carry out the binary image segmentation based on Renyi entropy, barrier zone in bianry image and background correctly can be distinguished, the region thus effectively removing barriers, is suitable to, for finding, the flat site reliable foundation of offer that aircraft security lands.

To achieve these goals, present invention employs techniques below scheme:

A kind of move and the image partition method of Renyi entropy based on gray scale, it is characterised in that it comprises the steps:

Step 1: the video flowing of aircraft collection is extracted image f, and (x, y), (x, the luminance component in y) carries out statistics of histogram based on this image f；

Step 2: this grey level histogram counted is carried out gray scale and moves；

Step 3: the grey level histogram of moving obtained after gray scale is moved carries out the binary image segmentation based on Renyi entropy, obtain correctly distinguishing barrier zone and background bianry image g (x, y).

The described grey level histogram described image f of reflection (x, the probability that in y), each gray value i occurs, i=0,1,2 ..., 255.

Described step 2 includes:

Step 2-1: in the described grey level histogram drawn out in described step 1, it is determined that the gray value MAX corresponding to probability peak occurred between gray value 100-200；

Step 2-2: if gray value MAX=150, then described grey level histogram translates 0 gray value；If 100≤MAX < 150, then described grey level histogram is to right translation (150-MAX) individual gray value；If 150 < MAX≤200, then described grey level histogram is to left (MAX-150) individual gray value；

Step 2-3: judge now more than 200 and less than or equal to whether the gray value in 255 scopes exists the probability more than 0: if existing, then by gray value (200,255] rectangular histogram picture corresponding in scope be perpendicular to abscissa center the longitudinal axis for axis of symmetry, mirror image moves gray value [0,55) in scope, if being absent from, then it is left intact.

Carry out including step based on the binary image segmentation of Renyi entropy to described grey level histogram of moving:

Step 3-1: determine optimal segmenting threshold t based on Renyi entropy^*；

Step 3-2: according to this optimal segmenting threshold t determined^*, to described step 1 obtains described image f (x, y) carries out binarization segmentation, obtain described bianry image g (x, y).

It is preferred that described optimal segmenting threshold t^*Determination step be:

Step 3-1-1: assume that (x, y) is divided into barrier zone O and background B two parts to segmentation threshold t, then the prior probability P of barrier zone O part by described image f_OThe prior probability P of (t) and background part B_BT () be following formula 2 respectively), 3):

$P_O\left(t\right)=\underset{i=0}{\overset{t}{\mathrm{\&Sigma;}}}{p}_i---2)$

$P_B\left(t\right)=\underset{i=t+1}{\overset{255}{\mathrm{\&Sigma;}}}{p}_i---3)$

And formula 2) and formula 3) meet formula P_O(t)+P_B(t)=1, wherein, i is gray value, i=0,1,2 ..., 255, p_iFor gray value i described image f (x, y) in occur probability, t is positive integer, 0 < t < 255；

Step 3-1-2: thus obtain corresponding to described image f (x, the one-dimensional Renyi entropy of barrier zone O part y)One-dimensional Renyi entropy with background part BRespectively such as following formula 4), 5) shown in:

$H_O^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}\ln \underset{i=0}{\overset{t}{\mathrm{\&Sigma;}}}{\left[\frac{p_i}{P_O\left(t\right)}\right]}^{\mathrm{\&alpha;}}---4)$

$H_B^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}\ln \underset{i=t+1}{\overset{255}{\mathrm{\&Sigma;}}}{\left[\frac{p_i}{1-P_O\left(t\right)}\right]}^{\mathrm{\&alpha;}}---5)$

Wherein, parameter alpha is arithmetic number, 0 < α < 1；

Step 3-1-3: described image f (x, population entropy H (t) y) is defined as following formula 6):

$H\left(t\right)=H_O^{\mathrm{\&alpha;}}\left(t\right)+H_B^{\mathrm{\&alpha;}}\left(t\right)---6),$

Therefore described optimal segmenting threshold t^*By following formula 7) obtain:

$t^{*}=\arg \underset{0<t<255}{\max }\left[H\left(t\right)\right]---7).$

Obtain after described image f (x, y) based on the following formula 8) binarization segmentation obtained in described step 1 described bianry image g (x, y):

$g(x,y)=\left\{\begin{array}{cc}0& f(x,y)\&le;t^{*}\\ 1& f(x,y)>t^{*}\end{array}\right.---8).$

The invention have the advantage that

The present invention grey level histogram to being counted by original image has carried out gray scale and has moved improvement, then the grey level histogram of moving obtained after gray scale being moved carries out the binary image segmentation based on Renyi entropy, to obtain correctly distinguishing the bianry image of barrier zone and background, the region thus can effectively remove barriers, is suitable to, for finding, the flat site reliable foundation of offer that aircraft security lands.The present invention is real-time, it is adaptable to different scenes, for instance the areas such as Gobi desert that physical features environment is complex or desert, it may be achieved the Accurate Segmentation to targets such as building, highway, waters.

Accompanying drawing explanation

Fig. 1 is the flowchart of the present invention.

Fig. 2 is the grey level histogram that original image example 1 is drawn out.

Fig. 3 be carry out grey level histogram shown in Fig. 2 obtaining after gray scale is moved move grey level histogram.

Fig. 4 is to moving, shown in Fig. 3, the bianry image obtained after grey level histogram carries out the binary image segmentation based on Renyi entropy.

Original image example 1 is drawn out grey level histogram by Fig. 5, does not carry out the bianry image obtained after gray scale is moved and is made directly the binary image segmentation based on Renyi entropy.

Fig. 6 is the grey level histogram that original image example 2 is drawn out.

Fig. 7 be carry out grey level histogram shown in Fig. 6 obtaining after gray scale is moved move grey level histogram.

Fig. 8 is to moving, shown in Fig. 7, the bianry image obtained after grey level histogram carries out the binary image segmentation based on Renyi entropy.

Original image example 2 is drawn out grey level histogram by Fig. 9, does not carry out the bianry image obtained after gray scale is moved and is made directly the binary image segmentation based on Renyi entropy.

Detailed description of the invention

As it is shown in figure 1, the image partition method that the present invention moves with Renyi entropy based on gray scale comprises the steps:

Step 1: the video flowing of aircraft collection is extracted image f (x, y), f (x, y) it is original image, usually, aircraft photographic head can start Real-time Collection at 3000 meters of of distance, image f (the x extracted from the video flowing gathered, y) it is stored in static memory with YCbCr format, wherein, Y-component is image f (x, y) luminance component, Cb, Cr component is image f (x respectively, y) blueness in, red concentration excursion amount composition, in practice, based on this image f (x, y) luminance component in carries out statistics of histogram, namely grey level histogram (grey level histogram herein is tradition grey level histogram) is drawn out；

Step 2: this grey level histogram counted is carried out gray scale and moves；

Step 3: the grey level histogram of moving obtained after gray scale is moved carries out the binary image segmentation based on Renyi entropy, obtain correctly distinguishing barrier zone and background bianry image g (x, y).

In reality is implemented, the reflection of the grey level histogram drawn out in step 1 be image f (x, the probability that in y), each gray value i occurs, i=0,1,2 ..., 255.

In reality is implemented, it is assumed that original image f (x, y) has M × N number of pixel, and image f (x, gray value set G={0 y), 1,2 ..., 255}, then Probability p_iBy following formula 1) obtain, thus drawing out grey level histogram, in grey level histogram, abscissa represents gray value i, i=0,1,2 ..., 255, vertical coordinate represents each gray value i Probability p occurred_i。

$p_i=\frac{n_i}{M\&times;N}---1)$

In formula 1) in, p_iFor gray value i at image f (x, the probability occurred in y), i ∈ G, n_iRepresent that gray value i is at image f (x, the number of times occurred in y).

In reality is implemented, step 2 comprises the steps that

Step 2-1: in the grey level histogram drawn out in step 1, it is determined that the gray value MAX corresponding to the probability peak that gray value occurs between 100-200；

Step 2-2: if gray value MAX=150, then grey level histogram integral translation 0 gray value；If 100≤MAX < 150, then grey level histogram is overall to right translation (150-MAX) individual gray value；If 150 < MAX≤200, then grey level histogram is overall to left (MAX-150) individual gray value；

Step 2-3: judge now more than 200 and less than or equal to whether the gray value in 255 scopes exists the probability more than 0, namely gray value is (200, 255] whether there is rectangular histogram picture in scope: if the gray value of the probability that existence is more than 0, then by gray value (200, 255] rectangular histogram picture corresponding in scope be perpendicular to abscissa center the longitudinal axis for axis of symmetry, mirror image (doubling) moves gray value [0, 55) in scope, gray value (200, 255] become in scope without any rectangular histogram picture, otherwise, if being absent from the gray value of the probability more than 0, then it is left intact.

In other words, in step 2, can making PC=MAX-150, then-50≤PC≤50, as MAX=150, PC=0, as 100≤MAX,<when 150, PC<0, when 150<during MAX≤200, PC>0.Span according to PC:

As PC=0, namely (x, during gray value MAX=150 corresponding to the probability peak between 100-200 of grey level histogram y), grey level histogram does not carry out gray scale and moves original image f, and namely grey level histogram does not change.

When PC is < when 0, i.e. original image f (x, y) the gray value MAX ∈ [100 that grey level histogram probability peak between 100-200 is corresponding, 150) time, grey level histogram moves in parallel to the right 150-MAX, what namely obtain after moving in parallel moves in grey level histogram, and original gray value MAX should correspond to gray value 150.

As PC > 0 time, i.e. original image f (x, y) the gray value MAX ∈ (150 that grey level histogram probability peak between 100-200 is corresponding, 200] time, grey level histogram moves in parallel MAX-150 to the left, what namely obtain after moving in parallel moves in grey level histogram, and original gray value MAX should correspond to gray value 150.

Such as, traditional grey level histogram shown in Fig. 2 is carried out gray scale move, because the gray value MAX corresponding to the probability peak of appearance between gray value 100-200 is approximately 130, it is positioned at the interval of 100 to 150, therefore, this grey level histogram entirety should move in parallel to the right 150-MAX, again because entirety moves in parallel to the right the grey level histogram after 150-MAX at gray value (200, 255] there is rectangular histogram picture (there is the probability more than 0) in scope, then by gray value (200, 255] the rectangular histogram picture in scope be perpendicular to abscissa center the longitudinal axis for axis of symmetry, mirror image moves gray value [0, 55) in scope, now gray value (200, 255] become in scope without any rectangular histogram picture, as shown in Figure 3.

In reality is implemented, carry out including step based on the binary image segmentation of Renyi entropy to moving grey level histogram:

Step 3-1: determine optimal segmenting threshold t based on Renyi entropy^*；

Step 3-2: according to this optimal segmenting threshold t determined^*, to step 1 obtains image f (x, y) carries out binarization segmentation, obtain bianry image g (x, y).

Wherein, optimal segmenting threshold t^*Determination step be:

Step 3-1-1: assume that (x, y) is divided into barrier zone O and background B two parts to segmentation threshold t, then the prior probability P of barrier zone O part by image f_OThe prior probability P of (t) and background part B_BT () be following formula 2 respectively), 3):

$P_O\left(t\right)=\underset{i=0}{\overset{t}{\mathrm{\&Sigma;}}}{p}_i---2)$

$P_B\left(t\right)=\underset{i=t+1}{\overset{255}{\mathrm{\&Sigma;}}}{p}_i---3)$

And formula 2) and formula 3) meet formula P_O(t)+P_B(t)=1, wherein, i is gray value, i=0,1,2 ..., 255, p_iFor gray value i image f (x, y) in occur probability, t is positive integer, 0 < t < 255；

Step 3-1-2: thus obtain corresponding to image f (x, the one-dimensional Renyi entropy of barrier zone O part y)One-dimensional Renyi entropy with background part BRespectively such as following formula 4), 5) shown in:

$H_O^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}\ln \underset{i=0}{\overset{t}{\mathrm{\&Sigma;}}}{\left[\frac{p_i}{P_O\left(t\right)}\right]}^{\mathrm{\&alpha;}}---4)$

$\begin{array}{c}{H}_B^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}\ln \underset{i=t+1}{\overset{255}{\mathrm{\&Sigma;}}}{\left[\frac{p_i}{P_B\left(t\right)}\right]}^{\mathrm{\&alpha;}}\\ =\frac{1}{1-\mathrm{\&alpha;}}\ln \underset{i=t+1}{\overset{255}{\mathrm{\&Sigma;}}}{\left[\frac{p_i}{1-P_O\left(t\right)}\right]}^{\mathrm{\&alpha;}}\end{array}---5)$

Wherein, parameter alpha is arithmetic number, 0 < α < 1；

Step 3-1-3: image f (x, population entropy H (t) y) is defined as following formula 6):

$H\left(t\right)=H_O^{\mathrm{\&alpha;}}\left(t\right)+H_B^{\mathrm{\&alpha;}}\left(t\right)---6),$

In formula 6) in, when α infinite tendency 1, it is shannon entropy,

Threshold value selection principle according to maximum entropy image segmentation, certain segmentation threshold t can make formula 6) obtaining maximum, then it is optimal segmenting threshold, therefore optimal segmenting threshold t^*By following formula 7) obtain:

$t^{*}=\arg \underset{0<t<255}{\max }\left[H\left(t\right)\right]---7).$

In actual applications, during the area such as the Gobi desert complex for physical features environment or desert, the binarization segmentation best results that α=0.6 obtains.

Reality implement in, obtain after image f (x, y) based on the following formula 8) binarization segmentation obtained in step 1 bianry image g (x, y):

$g(x,y)=\left\{\begin{array}{cc}0& f(x,y)\&le;t^{*}\\ 1& f(x,y)>t^{*}\end{array}\right.---8).$

Citing 1:

For desert area, if not adopting the present invention, on the basis of traditional grey level histogram (as shown in Figure 2) that original image is drawn out, not carrying out gray scale and move and be made directly the binary image segmentation based on Renyi entropy, the bianry image obtained is as shown in Figure 5.In Figure 5, barrier zone (obstacle such as highway) has been divided into black, and background (desert) has been divided into white.

But according to the present invention, on the basis of traditional grey level histogram (as shown in Figure 2) that original image is drawn out, carry out gray scale to move and obtain moving grey level histogram shown in Fig. 3, then again the grey level histogram of moving shown in Fig. 3 is carried out the binary image segmentation based on Renyi entropy, the bianry image obtained as shown in Figure 4, the optimal segmenting threshold t wherein tried to achieve^*It is 122.In the diagram, barrier zone (obstacle such as highway) has been divided into black, and background (desert) has been divided into white, and, in Fig. 5, more unshowned barrier zones also achieve certain performance with black in the diagram.

Citing 2:

For having the environment of forest etc., if not adopting the present invention, on the basis of traditional grey level histogram (as shown in Figure 6) that original image is drawn out, not carrying out gray scale and move and be made directly the binary image segmentation based on Renyi entropy, the bianry image obtained is as shown in Figure 9.In fig .9, barrier zone (obstacle such as building) has been divided into white, and background (forest etc.) has been divided into black.

But according to the present invention, on the basis of traditional grey level histogram (as shown in Figure 6) that original image is drawn out, carry out gray scale to move and obtain moving grey level histogram shown in Fig. 7, then again the grey level histogram of moving shown in Fig. 7 is carried out the binary image segmentation based on Renyi entropy, the bianry image obtained as shown in Figure 8, the optimal segmenting threshold t wherein tried to achieve^*It is 100.In fig. 8, barrier zone obstacles such as () buildings has been divided into black, and background (forest etc.) has been divided into white, achieves consistent with citing 1 to the segmentation of barrier Yu background.

Be can be seen that by the example above, for tradition grey level histogram, owing to aircraft is in flight course, difference along with environment, the monochrome information that target presents in the picture is different, as in desert area, the gray value in desert (background) is bigger than normal, the gray value of building is less than normal, therefore in tradition grey level histogram, region corresponding to probability peak is desert, namely Regional Representative's background that probability peak is corresponding, but under the environment with forest etc., the gray value of forest (background) is less than normal, the gray value of building is bigger than normal, therefore in tradition grey level histogram, region corresponding to probability peak is building, namely Regional Representative's barrier zone that probability peak is corresponding.Thus, tradition grey level histogram is after carrying out binary image segmentation, barrier zone is likely divided into white, it is also possible to be divided into black, visible, barrier zone and background cannot correctly be distinguished by the bianry image obtained after segmentation, thus bringing very big difficulty to the judgement of follow-up flat site.

And for the present invention, moving to process to the gray scale of tradition grey level histogram to make the gray value of the obstacle under different scene all reduce, so that the barrier zone in the bianry image obtained after carrying out the binary image segmentation based on Renyi entropy can be divided into 0 state (black) all the time, background is divided into 1 state (white) all the time, thus can realize the correct differentiation to barrier zone with background in the bianry image obtained after singulation.

In the present invention, from the video image of YCbCr format extract light intensity level, tradition grey level histogram drafting, be technology known in the art or prior art based on the cutting procedure of the binary image of Renyi entropy, therefore here do not describe in detail.

In actual applications, (x y) carries out RGB segmentation again and block-based can differentiate docks, namely can determine that and obtain being suitable to the flat site that aircraft security lands for bianry image g that the present invention is finally tried to achieve.

The present invention grey level histogram to being counted by original image has carried out gray scale and has moved improvement, then the grey level histogram of moving obtained after gray scale being moved carries out the binary image segmentation based on Renyi entropy, to obtain correctly distinguishing the bianry image of barrier zone and background, the region thus can effectively remove barriers, is suitable to, for finding, the flat site reliable foundation of offer that aircraft security lands.The present invention is real-time, it is adaptable to different scenes, for instance the areas such as Gobi desert that physical features environment is complex or desert, it may be achieved the Accurate Segmentation to targets such as building, highway, waters.

The above is presently preferred embodiments of the present invention and the know-why used thereof; for a person skilled in the art; when without departing substantially from the spirit and scope of the present invention; any based on apparent changes such as the equivalent transformation on technical solution of the present invention basis, simple replacements, belong within scope.

Claims (6)

1. move and the image partition method of Renyi entropy based on gray scale for one kind, it is characterised in that it comprises the steps:

Step 1: the video flowing of aircraft collection is extracted image f, and (x, y), (x, the luminance component in y) carries out statistics of histogram based on this image f；

Step 2: this grey level histogram counted is carried out gray scale and moves；

Step 3: the grey level histogram of moving obtained after gray scale is moved carries out the binary image segmentation based on Renyi entropy, obtain correctly distinguishing barrier zone and background bianry image g (x, y)；

Wherein, step 2 includes:

Step 2-1: in the grey level histogram drawn out in step 1, it is determined that the gray value MAX corresponding to probability peak occurred between gray value 100-200；

Step 2-2: if gray value MAX=150, then grey level histogram translates 0 gray value；If 100≤MAX < 150, then grey level histogram is to right translation (150-MAX) individual gray value；If 150 < MAX≤200, then grey level histogram is to left (MAX-150) individual gray value；

Step 2-3: judge now more than 200 and less than or equal to whether the gray value in 255 scopes exists the probability more than 0: if existing, then by gray value (200,255] rectangular histogram picture corresponding in scope be perpendicular to abscissa center the longitudinal axis for axis of symmetry, mirror image moves gray value [0,55) in scope, if being absent from, then it is left intact.

2. image partition method as claimed in claim 1, it is characterised in that:

The described grey level histogram described image f of reflection (x, the probability that in y), each gray value i occurs, i=0,1,2 ..., 255.

3. image partition method as claimed in claim 1, it is characterised in that:

Carry out including step based on the binary image segmentation of Renyi entropy to described grey level histogram of moving:

Step 3-1: determine optimal segmenting threshold t based on Renyi entropy^*；

Step 3-2: according to this optimal segmenting threshold t determined^*, to described step 1 obtains described image f (x, y) carries out binarization segmentation, obtain described bianry image g (x, y).

4. image partition method as claimed in claim 3, it is characterised in that:

Described optimal segmenting threshold t^*Determination step be:

Step 3-1-1: assume that (x, y) is divided into barrier zone O and background B two parts to segmentation threshold t, then the prior probability P of barrier zone O part by described image f_OThe prior probability P of (t) and background part B_BT () be following formula 2 respectively), 3):

$P_O\left(t\right)=\underset{i=0}{\overset{t}{\&Sigma;}}{p}_i---2)$

$P_B\left(t\right)=\underset{i=t+1}{\overset{255}{\&Sigma;}}{p}_i---3)$

And formula 2) and formula 3) meet formula P_O(t)+P_B(t)=1, wherein, i is gray value, i=0,1,2 ..., 255, p_iFor gray value i, at described image f, (x, the probability occurred in y), t is positive integer, 0 < t < 255；

Step 3-1-2: thus obtain corresponding to described image f (x, the one-dimensional Renyi entropy of barrier zone O part y)One-dimensional Renyi entropy with background part BRespectively such as following formula 4), 5) shown in:

$H_O^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}ln\underset{i=0}{\overset{t}{\&Sigma;}}{\&lsqb;\frac{p_i}{P_O\left(t\right)}\&rsqb;}^{\mathrm{\&alpha;}}---4)$

$H_B^{\mathrm{\&alpha;}}\left(t\right)=\frac{1}{1-\mathrm{\&alpha;}}ln\underset{i=t+1}{\overset{255}{\&Sigma;}}{\&lsqb;\frac{p_i}{1-P_O\left(t\right)}\&rsqb;}^{\mathrm{\&alpha;}}---5)$

Wherein, parameter alpha is arithmetic number, 0 < α < 1；

Step 3-1-3: described image f (x, population entropy H (t) y) is defined as following formula 6):

$H\left(t\right)=H_O^{\mathrm{\&alpha;}}\left(t\right)+H_B^{\mathrm{\&alpha;}}\left(t\right)---6),$

Therefore described optimal segmenting threshold t^*By following formula 7) obtain:

$t^{*}=\arg \underset{0<t<255}{max}\&lsqb;H\left(t\right)\&rsqb;---7).$

5. image partition method as claimed in claim 4, it is characterised in that:

Described parameter alpha takes 0.6.

6. image partition method as claimed in claim 3, it is characterised in that:

Obtain after described image f (x, y) based on the following formula 8) binarization segmentation obtained in described step 1 described bianry image g (x, y):

$g(x,y)=\left\{\begin{array}{cc}0& f(x,y)\&le;t^{*}\\ 1& f(x,y)>t^{*}\end{array}\right.---8).$