CN103325102A - Method for processing greasy weather degraded image based on histogram equalization - Google Patents

Abstract

The invention relates to a method for processing a greasy weather degraded image based on histogram equalization. The method comprises the steps that (1) the sky area in the greasy weather degraded image is separated, the grey value of dots in the greasy weather degraded image is changed to be the grey value of the sky area, and a primary processing image is obtained; (2) the method of histogram equalization is used for changing the primary processing image, and a novel image after being processed is obtained. The method removes the enhanced interference to effective information and the influence on lower end grey values of the sky area in an original method of histogram equalization, the degree of distortion of the image after being enhanced is lowered, and enhancement to the greasy weather degraded image is quickly and effectively achieved.

Description

Greasy weather degraded image disposal route based on histogram equalization

Technical field

The present invention relates to the processing of greasy weather degraded image, refer to a kind of greasy weather degraded image disposal route based on histogram equalization particularly.

Background technology

As a kind of mist of being classified as one of ten big hazard weathers in the world, to urban and rural highway transportation, aviation and navigation, electric system, industrial and agricultural production and daily life and even the healthy influence that all can produce in various degree, wherein mist is the most serious to the influence of road traffic transportation.In the road traffic transportation, mist can form serious dysopia to the driver, and causes one of major reason of traffic hazard during dysopia.Therefore, the how generation of pernicious road traffic accident under the visuality that improves the road environment system under the severe weather conditions such as greasy weather, prevention low visibility weather condition of research, become a research focus of traffic and message area, thereupon, how research is to having very important theory and practical significance too because the image degradation that severe weather conditions such as mist cause is effectively handled.

Processing to degraded image is divided into two big classes from big type, and the first kind is the figure image intensifying, and second class is image restoration.Wherein, image recovery method mainly depends on some information beyond the image, according to the former true picture of certain model and theoretical method recovery degraded image, for example RETINEX model, atmosphere degradation model, full variation model etc.And image enchancing method mainly is divided into two kinds of methods: 1) thus by conversion the contrast of image is stretched and to reach the requirement that improves image definition; 2) texture that strengthens to change image by the subspace with image reaches the effect of figure image intensifying.Traffic scene vision for the greasy weather strengthens the field, because its real-time requires high, therefore compared to the complexity theory model of image recovery method and fairly large amount of calculation, the amount of calculation of image enchancing method is little, treatment effect is excellent, makes it that fairly large application arranged in this regard.

At present, general employing histogram equalization method realizes the enhancing to image, this method makes the mean value of the probability that probability that the gray-scale value of adjacent point-to-point transmission occurs occurs greater than all gray-scale values, make contrast be enhanced, namely occurs the most for a long time in image when certain gray-scale value, its missionary society's stretching with adjacent gray-scale value is maximum.On the integral image effect, this method can make the overall contrast of image obtain maximum tension.

But the purpose of figure image intensifying is to make in the image Useful Information obtain maximum enhancing, rather than with occurrence number at most as unique enhancing standard.In the process that the greasy weather degraded image is handled, the gray-scale value of image sky dummy section is more approaching, and the frequency that occurs is very high, if directly adopt the histogram equalization method to strengthen, day dummy section then can occur because strengthening excessively distortion.And for Misty Image this class specific question of degenerating, the degree of degeneration of the different depth of field is inequality, but the scenery of the different depth of field, its gray-scale value fully may be identical or close.Therefore, for this type of problem, then can not be decided the amplitude of contrast stretching by the frequency that the gradation of image value occurs.

Summary of the invention

In view of above-mentioned the deficiencies in the prior art, the invention provides a kind of greasy weather degraded image disposal route based on histogram equalization, this method is at first separated a day dummy section, and the zone strengthens amplitude compression on high; The depth of field of each scene point in the estimated image then, stretch range and depth of field distance at the scene point place are approximated to proportional relation, thereby realize fast and effeciently the enhancing to the greasy weather degraded image.

Realize that the technical scheme that the object of the invention adopts is: a kind of greasy weather degraded image disposal route based on histogram equalization comprises:

(1) separates sky dummy section in the greasy weather degraded image, the gray-scale value of described greasy weather degraded image mid point is become the described day gray-scale value in the dummy section, obtain the rough handling image;

(2) adopt the described rough handling image of histogram equalization method conversion, the new images after obtaining handling.

In technique scheme, the sky dummy section in the described separation greasy weather degraded image may further comprise the steps:

(1-1) ask the gray-scale value Density Distribution of described greasy weather degraded image: q ₀, q ₁..., q ₂₅₅

(1-2) basis

Ask the gray-scale value probability distribution p of described greasy weather degraded image ₀, p ₁..., p ₂₅₅

(1-3) the sky dummy section of asking described greasy weather degraded image to divide based on density

Wherein

R is (i, j) gray values of pixel points;

(1-4) ask the gray-scale value Density Distribution q of described greasy weather degraded image ₀, q ₁..., q ₂₅₅Mathematical expectation: $E_q=\underset{i=0}{\overset{255}{\mathrm{\Σ}}}i{q}_i=i_0$

(1-5) with i ₀As the Preliminary division threshold value of sky dummy section, obtain the Density Distribution of approximate day dummy section:

,

(1-6) ask the mathematical expectation of described approximate day dummy section

And variances sigma _s

(1-7) the sky dummy section of asking described greasy weather degraded image to divide based on the mathematical expectation threshold value

Wherein

R is (i, j) gray values of pixel points;

(1-8) ask two kinds of sky dummy section S ₁And S ₂Intersection s _Ij, with s _IjAs the sky dummy section of last division, namely

$s_{\mathrm{ij}}=s_{\mathrm{ij}}^1^s_{\mathrm{ij}}^2$

(1-9) gray-scale value of original greasy weather degraded image mid point is become the sky dummy section s of described last division _IjIn gray-scale value, obtain the rough handling image.

Further, the described rough handling image of described employing histogram equalization method conversion may further comprise the steps:

(2-1) ask the gray-scale value Density Distribution q' of described rough handling image ₀, q' ₁..., q' ₂₅₅Maximum non-0 gray-scale value max _{Q '}With the non-0 gray-scale value min of minimum _{Q '}

(2-2) get $p_1^{\′}=q_{\mathrm{mi}{n}_{q^{\′}}}^{\′},$ $p_2^{\′}=q_{{\min }_{q^{\′}}+1}^{\′},$ …, $p_{r_0}^{\′}=q_{{\min }_{q^{\′}}+r_0-1}^{\′},$ $r_0={\max }_{q^{\′}}-2\sqrt{2}{\mathrm{\σ}}_s^{\′}-{\min }_{q^{\′}}+1,$ With vector $(p_1^{\′},p_2^{\′},\·\·\·,p_{r_0}^{\′})$ Normalization obtains $(\stackrel{\‾}{p_1^{\′}},\stackrel{\‾}{p_2^{\′}},...,\stackrel{\‾}{p_{r_0}^{\′}})=(p_1^{\′},p_{r_0}^{\′},...,p_{r_0}^{\′})/\underset{k=1}{\overset{r_0}{\mathrm{\Σ}}}{p}_k^{\′};$

(2-3) to arbitrary pixel of described rough handling image, establishing its gray-scale value is r, the gray-scale value Density Distribution after described rough handling image is obtained handling after by following transform conversion, and the new images after namely obtaining handling:

$s=T\left(r\right)=\left\{\begin{array}{cc}255,& r>\stackrel{\&OverBar;}{E_s^{\&prime;}}\\ 0.5*r+255-0.5*(\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;})& \stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ 0.5*{\min }_{q^{\&prime;}}+(255-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}-0.5*{\min }_{q^{\&prime;}})\underset{k=1}{\overset{r}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{p_k^{\&prime;}},& {\min }_{q^{\&prime;}}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ r/2,& {\min }_{q^{\&prime;}}<r\end{array}\right.$

The inventive method has been improved in original histogram equalization method interference that day dummy section strengthens effective information and the influence of low side gray-scale value, and image is being reduced through the degree of distortion after strengthening.In addition, the present invention adopts method of overall importance, and amount of calculation is less, is easy to use.

Description of drawings

Fig. 1 is the process flow diagram that the present invention is based on the greasy weather degraded image disposal route of histogram equalization;

Fig. 2 obtains the process flow diagram of rough handling image for day dummy section in the separation greasy weather degraded image;

Fig. 3 a is that first image adopts the image after original common histogram equalization method is handled;

Fig. 3 b is for adopting the image after the inventive method is handled first image;

Fig. 4 a is that second image adopts the image after original common histogram equalization method is handled;

Fig. 4 b is for adopting the image after the inventive method is handled second image.

Embodiment

The present invention is described in further detail below in conjunction with the drawings and specific embodiments.

As depicted in figs. 1 and 2, the greasy weather degraded image disposal route that the present invention is based on histogram equalization may further comprise the steps:

Step S100 separates the sky dummy section in the greasy weather degraded image, and the gray-scale value of described greasy weather degraded image mid point is become the described day gray-scale value in the dummy section, obtains the rough handling image;

Step S200 adopts the described rough handling image of improved histogram equalization method conversion, the new images after obtaining handling.

Because a day dummy section has following three key characters:

The gray-scale value of A, day dummy section is bigger, and it is histogrammic high-end to be in Misty Image, and approximate Normal Distribution.

B, day dummy section are using the contrast before the histogram equalization method strengthens smaller, within namely sky area grayscale value is in more among a small circle.

C, day dummy section be in the contrast increasing degree maximum of using after the histogram equalization method strengthens, and main cause is that the frequency that occurs of the gray-scale value of this part is higher, presses histogram equalizing method, gray-scale value increasing degree maximum wherein.

For this reason, among the step S100 of present embodiment, according to following quick sky territorial classification based on the histogram equalization method:

S101, ask the gray-scale value Density Distribution of original greasy weather degraded image I: q ₀, q ₁..., q ₂₅₅

S102, basis I is value 0,1,2 successively ... 255.Try to achieve the gray-scale value probability distribution p of original greasy weather degraded image I ₀, p ₁..., p ₂₅₅

S103, the sky dummy section of asking original greasy weather degraded image I to divide based on density Wherein

In the following formula, r is (i, gray-scale value j), the p of pixel in the image I _rBe point (i, gray-scale value probability distribution j).

S104, ask the gray-scale value Density Distribution q of original greasy weather degraded image I according to following formula ₀, q ₁..., q ₂₅₅Mathematical expectation E _q:

$E_q=\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}{\mathrm{iq}}_i=i_0$

S105, with i ₀As the Preliminary division threshold value of sky dummy section, obtain the Density Distribution of approximate day dummy section:

,

S106, ask the mathematical expectation of described approximate day dummy section by following formula respectively

And variances sigma _s:

$\stackrel{\&OverBar;}{E}s=\underset{i=i_0}{\overset{255}{\mathrm{\&Sigma;}}}i{\stackrel{\&OverBar;}{q}}_i,$

${\mathrm{\&sigma;}}_s=E\left(i^2\right)-\stackrel{\&OverBar;}{E_s^2}$

In the following formula, i=i ₀, i ₁..., i ₂₅₅

S107, the sky dummy section of asking original greasy weather degraded image I to divide based on the mathematical expectation threshold value

Wherein

In the following formula, r is (i, gray-scale value j), the σ of pixel in the image I _sTried to achieve by S106.

S108, ask two kinds of sky dummy section S ₁And S ₂Intersection s _Ij, with s _IjAs the sky dummy section of last division, namely

$s_{\mathrm{ij}}=s_{\mathrm{ij}}^1^s_{\mathrm{ij}}^2$

S109, the gray-scale value of original greasy weather degraded image I mid point become the sky dummy section s of last division _IjIn gray-scale value, obtain rough handling image II.

At the grey value profile low side of image, the minimum gradation value of establishing this image is q ₀, the histogram equalization method is with q ₀Following gray-scale value all is transformed to 0, i.e. black.Therefore, increase by a linear transitions interval herein, make conversion continuous, the effective information of image is kept simultaneously.Gray-scale value at image is high-end, be mainly a day dummy section, the contrast of this subregion does not need to increase, and adopts the compression-type linear transformation, this regional pixel the most significant end of gray-scale value after the conversion among a small circle within, in order to vacate the stretching that more space is used for effective information.

So present embodiment step S200 utilizes improved histogram equalization method to handle the rough handling image II that above-mentioned steps S109 obtains, concrete steps are as follows:

S201, ask the gray-scale value Density Distribution q' of image II ₀, q' ₁..., q' ₂₅₅, obtain maximum non-0 gray-scale value

Non-0 gray-scale value with minimum

, and mathematical expectation

With i' ₀As the Preliminary division threshold value of sky dummy section, obtain the Density Distribution of approximate day dummy section:

,

And obtain the mathematical expectation of the approximate day dummy section of this distribution

And variances sigma ' _s

S202, get $p_1^{\&prime;}=q_{\mathrm{mi}{n}_{q^{\&prime;}}}^{\&prime;},$ $p_2^{\&prime;}=q_{{\min }_{q^{\&prime;}}+1}^{\&prime;},$ …, $p_{r_0}^{\&prime;}=q_{{\min }_{q^{\&prime;}}{+r}_0-1}^{\&prime;},$ $r_0={\max }_{q^{\&prime;}}-2\sqrt{2}{\mathrm{\&sigma;}}_s^{\&prime;}-{\min }_{q^{\&prime;}}+1,$ With vector $(p_1^{\&prime;},p_2^{\&prime;},\&CenterDot;\&CenterDot;\&CenterDot;,p_{r_0}^{\&prime;})$ Normalization obtains $(\stackrel{\&OverBar;}{p_1^{\&prime;}},\stackrel{\&OverBar;}{p_2^{\&prime;}},...,\stackrel{\&OverBar;}{p_{r_0}^{\&prime;}})=(p_1^{\&prime;},p_{r_0}^{\&prime;},...,p_{r_0}^{\&prime;})/\underset{k=1}{\overset{r_0}{\mathrm{\&Sigma;}}}{p}_k^{\&prime;}.$

S203, to arbitrary pixel of rough handling image II (i, j), establishing its gray-scale value is r, carries out conversion according to transform:

$s=T\left(r\right)=\left\{\begin{array}{cc}255,& r>\stackrel{\&OverBar;}{E_s^{\&prime;}}\\ 0.5*r+255-0.5*(\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;})& \stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ 0.5*{\min }_{q^{\&prime;}}+(255-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}-0.5*{\min }_{q^{\&prime;}})\underset{k=1}{\overset{r}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{p_k^{\&prime;}},& {\min }_{q^{\&prime;}}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ r/2,& {\min }_{q^{\&prime;}}<r\end{array}\right.$

Obtain handling gray-scale value Density Distribution later by this conversion, the new images III after namely the image II obtains handling after by the histogram equalization method conversion after improving.

The result can see by experiment, and Fig. 3 a, Fig. 4 a are for adopting the image after original common histogram equalization method is handled, and effective fiducial interval of common histogram equalization method originally is

σ is the variance of original image I, and at gradation of image value low side, having the part effective information becomes black region; Fig. 3 b, Fig. 4 b are the image after the inventive method is handled, and between the effective information drawing zone of the improved histogram equalization method of the present invention are

σ _sIt is the variance yields of image II; have brighter image information, and compared to common histogram equalization method, improved method is protected to the relevant information of low side; be that main information has been carried out translation and compression to high-end sky dummy section, make the degree of distortion after the figure image intensifying littler.

Claims (3)

1. the greasy weather degraded image disposal route based on histogram equalization is characterized in that, comprising:

(1) separates sky dummy section in the greasy weather degraded image, the gray-scale value of described greasy weather degraded image mid point is become the described day gray-scale value in the dummy section, obtain the rough handling image;

(2) adopt the described rough handling image of histogram equalization method conversion, the new images after obtaining handling.

2. according to the described greasy weather degraded image disposal route based on histogram equalization of claim 1, it is characterized in that the sky dummy section in the described separation greasy weather degraded image may further comprise the steps:

(1-1) ask the gray-scale value Density Distribution of described greasy weather degraded image: q ₀, q ₁..., q ₂₅₅

(1-2) basis Ask the gray-scale value probability distribution p of described greasy weather degraded image ₀, p ₁..., p ₂₅₅

(1-3) the sky dummy section of asking described greasy weather degraded image to divide based on density Wherein

R is (i, j) gray values of pixel points;

(1-4) ask the gray-scale value Density Distribution q of described greasy weather degraded image ₀, q ₁..., q ₂₅₅Mathematical expectation: $E_q=\underset{i=0}{\overset{255}{\mathrm{\&Sigma;}}}i{q}_i=i_0;$

(1-5) with i ₀As the Preliminary division threshold value of sky dummy section, obtain the Density Distribution of approximate day dummy section:

,

(1-6) ask the mathematical expectation of described approximate day dummy section

And variances sigma _s

(1-7) the sky dummy section of asking described greasy weather degraded image to divide based on the mathematical expectation threshold value

Wherein

R is (i, j) gray values of pixel points;

(1-8) ask two kinds of sky dummy section S ₁And S ₂Intersection s _Ij, with s _IjAs the sky dummy section of last division, namely

$s_{\mathrm{ij}}=s_{\mathrm{ij}}^1^s_{\mathrm{ij}}^2$

(1-9) gray-scale value of original greasy weather degraded image mid point is become the sky dummy section s of described last division _IjIn gray-scale value, obtain the rough handling image.

3. according to the described greasy weather degraded image disposal route based on histogram equalization of claim 2, it is characterized in that the described rough handling image of described employing histogram equalization method conversion may further comprise the steps:

(2-1) ask the gray-scale value Density Distribution q' of described rough handling image ₀, q' ₁..., q' ₂₅₅Maximum non-0 gray-scale value

With non-0 gray-scale value of minimum

(2-2) get $p_1^{\&prime;}=q_{\mathrm{mi}{n}_{q^{\&prime;}}}^{\&prime;},$ $p_2^{\&prime;}=q_{{\min }_{q^{\&prime;}}+1}^{\&prime;},$ ...， $p_{r_0}^{\&prime;}=q_{{\min }_{q^{\&prime;}}+r_0-1}^{\&prime;},$ $r_0={\max }_{q^{\&prime;}}-2\sqrt{2}{\mathrm{\&sigma;}}_s^{\&prime;}-{\min }_{q^{\&prime;}}+1,$ With vector $(p_1^{\&prime;},p_2^{\&prime;},\&CenterDot;\&CenterDot;\&CenterDot;,p_{r_0}^{\&prime;})$ Normalization obtains $(\stackrel{\&OverBar;}{p_1^{\&prime;}},\stackrel{\&OverBar;}{p_2^{\&prime;}},...,\stackrel{\&OverBar;}{p_{r_0}^{\&prime;}})=(p_1^{\&prime;},p_{r_0}^{\&prime;},...,p_{r_0}^{\&prime;})/\underset{k=1}{\overset{r_0}{\mathrm{\&Sigma;}}}{p}_k^{\&prime;};$

(2-3) to arbitrary pixel of described rough handling image, establishing its gray-scale value is r, the gray-scale value Density Distribution after described rough handling image is obtained handling after by following transform conversion, and the new images after namely obtaining handling:

$s=T\left(r\right)=\left\{\begin{array}{cc}255,& r>\stackrel{\&OverBar;}{E_s^{\&prime;}}\\ 0.5*r+255-0.5*(\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;})& \stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}+\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ 0.5*{\min }_{q^{\&prime;}}+(255-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}-0.5*{\min }_{q^{\&prime;}})\underset{k=1}{\overset{r}{\mathrm{\&Sigma;}}}\stackrel{\&OverBar;}{p_k^{\&prime;}},& {\min }_{q^{\&prime;}}<r\&le;\stackrel{\&OverBar;}{E^{\&prime;}}-\sqrt{2}{\mathrm{\&sigma;}}^{\&prime;}\\ r/2,& {\min }_{q^{\&prime;}}<r\end{array}\right..$

Images