CN104867115A - Color cast detection and image enhancement method for images after defogging - Google Patents

Abstract

The invention relates to a color cast detection and image enhancement method for images after defogging. Brightness radiation estimation is performed on the images after defogging by utilizing a guiding filter with low complexity and the edge preservation characteristic so that color cast of the images after defogging is corrected and the images are enhanced. Firstly, color cast detection is performed on the images; then a gray world method is performed on the images exceeding the color cast threshold value; and an improved single-scale Retinex method is performed on the images. With application of the method, color cast detection and image enhancement of the images after defogging are realized.

Description

After a kind of mist elimination, image colour cast detects and image enchancing method

To the present invention relates in image mist elimination image color chips after a kind of mist elimination to detect and image enchancing method, belong to image mist elimination field.

Background technology

At present be the society of information-based develop rapidly, there occurs qualitative leap at the acquisition of information, the everyway such as process and propagation, image information can account for about 60% of gross information content.Image is the basis of vision, and people utilize image to record, obtain, propagate important information, and it can show all kinds of personages and the events such as natural landscape, humane landscape, daily routines.Image uses all kinds of imaging device, in several ways and angle views obtain, direct impact formed to human eye, obtain the entity of visual perception.The quality of image also drastically influence reading and the judgement of information.

The image obtained from image acquisition device is difficult to meet application demand due to a variety of causes, needs to process image, convert the quality improving image, reaches the demand meeting human vision and psychology while application requires.Digital Image Processing is that certain width picture is become digital form from original formal transformation, forms the image of numeral through digitizing.Digital image processing techniques appear at the initial stage in last century the earliest, and this field receives great attention in decades, have developed into an independent educational project.

Along with the aggravation of environmental pollution, air quality serious degradation, fog weather is also more and more frequent, and PM2.5 value also becomes the topic that people more and more pay close attention to.In fog weather, because dust granules has the effect of scattering and absorption to light, thus light intensity is weakened, the light intensity entering receiving equipment changes, thus reduce the contrast of image, result in image blurring, that sharpness decays, reduces image resolution.Meanwhile, also easily there is the phenomenon of colour cast in image, does not meet human vision effect.These visually ambiguous picture information is lost to a certain extent, have impact on the accuracy of result of determination, constrain the performance of the effectiveness such as identification and tracking, navigational system, supervisory system, military observation system of target, serious impact is all brought on the production of society, the life of people.Such as freeway monitoring system, fog weather reduces atmospheric visibility, and driver can be inaccurate to the judgement of surrounding traffic situation, extremely easily traffic hazard occurs in this case.

In order to obtain the second best in quality image at fog weather, a lot of scholar begins one's study the problem of image mist elimination.Image mist elimination is through various method and is removed the impact that image produces by fog, restores image that is clear, lively, that meet human visual system.Image mist elimination technology has interdisciplinary feature, has a extensive future, and is one of research direction of image processing field, receives the concern of domestic and international numerous researchers, has become the study hotspot of computer vision and image processing field.Image mist elimination, as a new branch of science, relates to randomness and the very large weather conditions of complicacy.Up to the present, although emerged in large numbers the method for a lot of image mist elimination.But for fog image, also can along with image distortion in various degree while removal mist, such as image colour cast, therefore needs to find effective method and carries out colour cast detection and image enhaucament to image after mist elimination.

Summary of the invention

The object of the invention is, propose image color chips after a kind of mist elimination and detect and image enchancing method, achieve the object optimizing image after mist elimination, there is applicability widely simultaneously.

Technical scheme of the present invention is as follows:

After a kind of mist elimination, image colour cast detects and image enchancing method, comprises the steps:

The first step, carries out color chips detection to the image after mist elimination;

Second step, performs gray world algorithm to the image exceeding colour cast threshold values;

3rd step, performs the single scale Retinex algorithm improved to image;

1, colour cast threshold values is chosen

Choosing colour cast threshold values is 1.5.If the colour cast K calculated _cast>=1.5, then process decision chart picture has colour cast, enters step 2, advanced Grey-world algo-rithms, then uses the single scale Retinex algorithm of improvement to strengthen image; If K _cast< 1.5, enters step 3, then think that image does not have colour cast situation, directly uses the single chi Retinex algorithm for image enhancement improved to the image enhancement processing after mist elimination.

2, the single scale Retinex image enchancing method improved

The rough estimate of A, brightness radiation;

Method of estimation is:

$L={\left(\underset{c\&Element;\{R,G,B\}}{\mathrm{mean}}\left(J(:,:,c)\right)\right)}^{{\mathrm{\&gamma;}}_1}---\left(1\right)$

Here, J, mean (J) are respectively the average corresponding with it of the image after mist elimination;

γ ₁=min (min (J)+0.6,1) is gamma factor.

B, utilize the accurate estimated brightness radiation of Steerable filter device;

Method of estimation is:

L _refine＝guidedfilter(J _gray,L,ω _l,ε _l) (2)

Here, guidedfilter () is the reduced form of Steerable filter device; J _grayfor guiding figure; L, ω _l, ε ₁; Be respectively thick illumination estimate, the size of filter window and penalty term.

The enhancing of image after C, mist elimination

According to the brightness radiation after accurate estimation, the image after mist elimination is strengthened.Its expression formula is as follows:

$J_{\mathrm{enhance}}(x,y,c)=\frac{\exp (\log \left(J(x,y,R)\right)-\mathrm{\&alpha;}\&CenterDot;\log \left(L_{\mathrm{refine}}(x,y)\right))}{J(x,y,R)}\&CenterDot;J(x,y,c)---\left(3\right)$

Here, c is color component, α=1/exp (L _refine(x, y)) be the brightness regulation factor, the object of introducing avoids the highlight regions of image too saturated after strengthening.As can be seen from formula (3), because each color component strengthens on year-on-year basis, the image after therefore strengthening is that brightness increases, and colourity does not change.

Accompanying drawing explanation

Fig. 1: after mist elimination of the present invention, image colour cast detects and image enhaucament process flow diagram.

Embodiment

Be described principle of the present invention and feature below in conjunction with accompanying drawing, example, only for explaining the present invention, is not intended to limit scope of the present invention.

Be that after giving mist elimination, image colour cast detects and image enhaucament process flow diagram shown in Fig. 1, and will realize detecting and image enhaucament image colour cast after mist elimination, following link is crucial:

(1) colour cast threshold values is chosen

Choosing colour cast threshold values is 1.5.If the colour cast K calculated _cast>=1.5, then process decision chart picture has colour cast, enters step 2, advanced Grey-world algo-rithms, then uses the single scale Retinex algorithm of improvement to strengthen image; If K _cast< 1.5, enters step 3, then think that image does not have colour cast situation, directly uses the single chi Retinex algorithm for image enhancement improved to the image enhancement processing after mist elimination.

(2) the single scale Retinex image enchancing method improved

The rough estimate of A, brightness radiation;

Method of estimation is:

$L={\left(\underset{c\&Element;\{R,G,B\}}{\mathrm{mean}}\left(J(:,:,c)\right)\right)}^{{\mathrm{\&gamma;}}_1}---\left(1\right)$

Here, J, mean (J) are respectively the average corresponding with it of the image after mist elimination;

γ ₁=min (min (J)+0.6,1) is gamma factor.

B, utilize the accurate estimated brightness radiation of Steerable filter device;

Method of estimation is:

L _refine＝guidedfilter(J _gray,L,ω _l,ε _l) (2)

Here, guidedfilter () is the reduced form of Steerable filter device; J _grayfor guiding figure; L, ω _l, ε ₁; Be respectively thick illumination estimate, the size of filter window and penalty term.

The enhancing of image after C, mist elimination

According to the brightness radiation after accurate estimation, the image after mist elimination is strengthened.Its expression formula is as follows:

$J_{\mathrm{enhance}}(x,y,c)=\frac{\exp (\log \left(J(x,y,R)\right)-\mathrm{\&alpha;}\&CenterDot;\log \left(L_{\mathrm{refine}}(x,y)\right))}{J(x,y,R)}\&CenterDot;J(x,y,c)---\left(3\right)$

Here, c is color component, α=1/exp (L _refine(x, y)) be the brightness regulation factor, the object of introducing avoids the highlight regions of image too saturated after strengthening.As can be seen from formula (3), because each color component strengthens on year-on-year basis, the image after therefore strengthening is that brightness increases, and colourity does not change.

Claims (2)

1. after mist elimination, image colour cast detects and an image enchancing method, it is characterized in that comprising the steps:

Step one, carries out color chips detection to the image after mist elimination;

Step 2, performs gray world method to the image exceeding colour cast threshold values;

Step 3, performs the single scale Retinex method improved to image.

2. after a kind of mist elimination according to claim 1, image colour cast detects and image enchancing method, it is characterized in that, in described step one and step 3:

(1) colour cast threshold values is chosen

Choosing colour cast threshold values is 1.5, if the colour cast K calculated _cast>=1.5, then process decision chart picture has colour cast, enters step 2, advanced grey-world method, then uses the single scale Retinex algorithm of improvement to strengthen image; If enter step 3, then think that image does not have colour cast situation, directly use the single chi Retinex algorithm for image enhancement improved to the image enhancement processing after mist elimination;

(2) the single scale Retinex image enchancing method improved

The rough estimate of A, brightness radiation;

$L={\left(\underset{c\&Element;\{R,G,B\}}{\mathrm{mean}}\left(J(:,:,c)\right)\right)}^{{\mathrm{\&gamma;}}_1}---\left(1\right)$

Here, J, mean (J) are respectively the average corresponding with it of the image after mist elimination;

γ ₁=min (min (J)+0.6,1) is gamma factor;

B, utilize the accurate estimated brightness radiation of Steerable filter device;

Method of estimation is:

L _refine＝guidedfilter(J _gray,L,ω _l,ε _l) (2)

Here, guidedfilter () is the reduced form of Steerable filter device; J _grayfor guiding figure; L, ω _l, ε ₁be respectively thick illumination estimate, the size of filter window and penalty term;

The enhancing of image after C, mist elimination

According to the brightness radiation after accurate estimation, strengthen the image after mist elimination, its expression formula is as follows:

$J_{\mathrm{enhance}}(x,y,c)=\frac{\exp (\log \left(J(x,y,R)\right)-\mathrm{\&alpha;}\&CenterDot;\log \left(L_{\mathrm{refine}}(x,y)\right))}{J(x,y,R)}\&CenterDot;(x,y,c)---\left(3\right)$

Here, c is color component, α=1/exp (L _refine(x, y)) be the brightness regulation factor.