CN104200435A - Single-image defogging method based on fog imaging model - Google Patents

Abstract

The invention relates to a single-image defogging method based on a fog imaging model. The single-image defogging method based on the fog imaging model includes steps that reading an original image with fog; setting the atmospheric optical color as a fixed value; assuming that the albedo of the local region of the image is a constant vector, and there is local statistic irrelevance between the object surface chroma and medium propagation, and using independent component analysis to estimate the constant vector albedo so as to obtain a transmittance and a scene depth; carrying out normalization treatment on the transmittance; substituting into a physical model according to the transmittance and atmospheric optical color parameter, and outputting a defogged clear image. The single-image defogging method based on the fog imaging model is capable of improving the problem of serious partial distortion of the defogged image through the transmittance normalization method and obtaining a reduced image with good contrast ratio.

Description

A kind of single image defogging method capable based on greasy weather imaging model

Technical field

The invention belongs to digital image processing techniques field, be specifically related to a kind of single image defogging method capable based on greasy weather imaging model.

Background technology

Picture contrast and resolution that under foggy environment, video camera is taken obviously decline, and make the scene image observed under foggy environment muddy, are unfavorable for the observation of human eye.Along with expanding economy, the fields such as video monitoring, target identification are increasing to the dependence of image, so image mist elimination is very necessary simultaneously.

Mist is when the relative humidity of air molecule reaches capacity, and some air molecules are condensed to little water droplet and produce, and the gas sol particles in these atmosphere can have scattering process to imaging light.The light losing that scattering causes can make " transmitted light " strength retrogression, thereby causes the contrast of image to decline.And that the unevenness of particulate can make image thicken is unclear.Meanwhile, particulate also can be because of the effect of Multiple Scattering to the scattered portion of light, and original forward scattering is partly superimposed together imaging, thereby produces certain fuzzy.The summation of above-mentioned various effects, causes last Misty Image can produce serious degeneration, and smudgy, contrast is on the low side, and the minutia of scenery is covered, the unsaturated and distortion of color.

At present, the method for raising Misty Image sharpness mainly contains two classes: non-modelling and the algorithm based on model.Non-model Misty Image mist elimination algorithm is not considered the material elements that image deterioration produces, but only from the angle of figure image intensifying, processes image, to reach the object that improves picture quality, simply says to be exactly the overall contrast that strengthens image.Common method has histogram equalization, homomorphic filtering and Retinex algorithm etc., this method is applied widely, can have the contrast that more effectively improves image, and the grain details of outstanding image, but may cause the information loss of outshot, make the image after processing have distortion.Defogging method capable based on mist model is mist image to be carried out to a contrary inverse process of rain imaging recover without mist image.The He Kaiming of Hong Kong Chinese University etc. has proposed a kind of mist elimination algorithm (reference papers < < Single Image Haze Removal Using Dark Channel Prior > >) based on dark primary priori, this algorithm effect is outstanding, but algorithm complex is very large, limited its widespread use in engineering.

Summary of the invention

The present invention solves the existing technical matters of prior art; Providing a kind of not only can have good image mist elimination effect, and can solve the single image defogging method capable based on greasy weather imaging model that computation process complicated and time consumption that the algorithm of prior art exists is difficult to be applied to the problem of practical matter.

Above-mentioned technical matters of the present invention is mainly solved by following technical proposals:

A single image defogging method capable based on greasy weather imaging model, is characterized in that, comprises the following steps:

Step 1: read in the original mist image I (x) that has;

Step 2: image mist elimination based on greasy weather imaging model I (x), the coordinate of putting on x presentation video, I (x) represents the image of observing, t (x) is the transmissivity that a scalar represents light, A represents the color of atmosphere light, J (x) true picture recovering of indicating;

I (x)=J (x) t (x) R+ (1-t (x)) A formula one

R is decomposed along both direction, is respectively A direction and perpendicular to A direction; Residual vector after A Directional Decomposition is represented with R':

$I\left(x\right)=t\left(x\right)l^{\&prime;}\left(x\right)(\frac{R^{\&prime;}}{\left|\right|R^{\&prime;}\left|\right|}+\mathrm{\&eta;}\frac{A}{\left|\right|A\left|\right|})+(1-t\left(x\right))A$ Formula two

Wherein, r is a three-dimensional RGB vector, presentation surface reflection coefficient, and R' represents that R is along the component of A Directional Decomposition; L (x) is a scalar, represents how much catoptrically from body surface, l'(x) represents that l (x) is along the component of A Directional Decomposition;

Wherein, transmissivity t (x) is based on following formula:

$t\left(x\right)=1-\frac{(I_A\left(x\right)-\mathrm{\&eta;}{I}_{R^{\&prime;}}\left(x\right))}{\left|\right|A\left|\right|}$ Formula three

Wherein, η is based on following formula:

$\mathrm{\&eta;}=\frac{C_{\mathrm{\&Omega;}}(I_A,h)}{C_{\mathrm{\&Omega;}}(I_{R^{\&prime;}},h)}$ Formula four

Wherein c _Ω(I _a, h) and C _Ω(I _r', h) represent respectively I _awith h and I _r'covariance with h;

Step 3: obtain transmissivity t (x) value based on formula four and formula three, and t (x) is normalized:

$t^{\&prime;}\left(x\right)=\frac{{t-t}_{\min }}{t_{\max }-t_{\min }}$ Formula five the transmissivity t'(x based on after adjusting) and formula one, according to formula, draw final J (x):

$J\left(x\right)=\frac{I\left(x\right)-\left(1-t\left(x\right)\right)\&times;A}{t^{\&prime;}\left(x\right)}$

Therefore, tool of the present invention has the following advantages: 1, can effectively process coloured image and also can process gray level image; 2, algorithm process effect is better, and the brighter part of light can both clearly show, and there is no distortion; 3, for asking transmissivity t (x), the normalized that the utilization of this algorithm is comparatively easy, can either effectively solve problem of dtmf distortion DTMF, makes clear picture, again can shortcut calculation, improve image processing speed.

Accompanying drawing explanation

Fig. 1 is method flow schematic diagram of the present invention.

Fig. 2 is the pending original mist image one that has in the embodiment of the present invention.

Fig. 3 be after processing in the embodiment of the present invention, obtain without mist image one.

Fig. 4 is the pending original mist image two that has in the embodiment of the present invention.

Fig. 5 is the result of a kind of image defogging method capable of non-use the present invention.

Fig. 6 is this method result.

Embodiment

Below by embodiment, and by reference to the accompanying drawings, technical scheme of the present invention is described in further detail.

Embodiment:

One, first, set forth the method applied in the present invention: the physical model of propagating in the greasy weather according to light, as follows to atmospheric scattering model description:

I(x)＝J(x)t(x)+(1-t(x))×A

The coordinate of putting on x presentation video, I (x) represents the image observe, and t (x) is a scalar, represents the transmissivity of light, and A represents the color of atmosphere light, J (x) true picture recovering of indicating.

To formula, rewrite as follows:

I(x)＝t(x)l(x)R+(1-t(x))A

J=R*l wherein, R is a three-dimensional RGB vector, presentation surface reflection coefficient; L is a scalar, represents how much catoptrical from body surface.

Step 1: will have mist image to read in, and be designated as I (x).

Step 2: think that atmosphere light color is certain value, in this example, suppose its value for [0.8,0.8,0.9] '

Step 3: R is decomposed along both direction, is respectively A direction and perpendicular to A direction.Residual vector after A Directional Decomposition is represented with R'.

Can obtain formula:

$I\left(x\right)=t\left(x\right)l\left(x\right)(R^{\&prime;}+|\left|R\right|\left|\cos \mathrm{\&theta;\&eta;}\frac{A}{\left|\right|A\left|\right|}\right)+(1-t\left(x\right))A$

Step 4: make l'(x)=l (x) R', has:

$I\left(x\right)=t\left(x\right)l^{\&prime;}\left(x\right)(\frac{R^{\&prime;}}{\left|\right|R\left|\right|}+\frac{\left|\right|R\left|\right|\left|\right|A\left|\right|\cos \mathrm{\&theta;}\frac{A}{\left|\right|A\left|\right|}}{\left|\right|R^{\&prime;}\left|\right|\left|\right|A\left|\right|})+(1-t\left(x\right))A$

Again $<\stackrel{\&RightArrow;}{a},\stackrel{\&RightArrow;}{b}>=\left|\stackrel{\&RightArrow;}{a}\right|\left|\stackrel{\&RightArrow;}{b}\right|\cos <\stackrel{\&RightArrow;}{a},\stackrel{\&RightArrow;}{b}>,$ Above formula becomes:

$I\left(x\right)=t\left(x\right)l^{\&prime;}\left(x\right)(\frac{R^{\&prime;}}{\left|\right|R^{\&prime;}\left|\right|}+\frac{<R,A>}{\left|\right|R^{\&prime;}\left|\right|\left|\right|A\left|\right|}\&CenterDot;\frac{A}{\left|\right|A\left|\right|})+(1-t\left(x\right))A$

Step 5: order $\mathrm{\&eta;}=\frac{<R,A>}{\left|\right|R^{\&prime;}\left|\right|\left|\right|A\left|\right|},$ Have:

$I\left(x\right)=t\left(x\right)l^{\&prime;}\left(x\right)(\frac{R^{\&prime;}}{\left|\right|R^{\&prime;}\left|\right|}+\mathrm{\&eta;}\frac{A}{\left|\right|A\left|\right|})+(1-t\left(x\right))A$

Step 6: I (x), also along A direction and vertical A Directional Decomposition, is obtained:

$I_A\left(x\right)=\frac{<I\left(x\right),A>}{\left|\right|A\left|\right|}=t\left(x\right)l^{\&prime;}\left(x\right)\mathrm{\&eta;}+(1-t\left(x\right))\left|\right|A\left|\right|$

Step 7: because I (x) ²=I _a(x) ²+ I _r'(x) ²,

${\mathrm{IR}}^{\&prime;}\left(x\right)=\sqrt{{\left|\right|I\left(x\right)\left|\right|}^2-I_A{\left(x\right)}^2}=t\left(x\right)l^{\&prime;}\left(x\right)$

:

$t\left(x\right)=1-\frac{(I_A\left(x\right)-\mathrm{\&eta;}{I}_{R^{\&prime;}}\left(x\right))}{\left|\right|A\left|\right|}$

Like this, mist elimination work has just become and has asked η.

Step 8: think that these two amounts of t and l are separate, covariance is 0.Consider $C_{\mathrm{\&Omega;}}(t,\frac{1}{l^{\&prime;}})=0$

${\left(l^{\&prime;}\left(x\right)\right)}^{-1}=(1-(I_A\left(x\right)-\mathrm{\&eta;}{I}_{R^{\&prime;}}\left(x\right))/|\left|A\right|\left|\right)/I_{R^{\&prime;}}\left(x\right)=\frac{1-\frac{I_A\left(x\right)}{\left|\right|A\left|\right|}}{I_{R^{\&prime;}}\left(x\right)}+\frac{\mathrm{\&eta;}}{\left|\right|A\left|\right|}$

Define a new variables h, and order

$h\left(x\right)=\frac{\left|\right|A\left|\right|-I_A}{I_{R^{\&prime;}}}$

Step 9::

$\mathrm{\&eta;}=\frac{C_{\mathrm{\&Omega;}}(I_A,h)}{C_{\mathrm{\&Omega;}}(I_{R^{\&prime;}},h)}$

We have just obtained the expression formula of η in normal albedo situation like this, and this is the form of a ratio, make the meaningful certain assurance C of η _Ω(I _r', be not h) 0.

Step 10: t (x) is normalized.

$t^{\&prime;}=\frac{{t-t}_{\min }}{t_{\max }-t_{\min }}$

Step 11: obtain after the expression formula of η, can draw final J (x) according to formula.

Two, Fig. 1 and Fig. 4 are the pending original mist image that has; Fig. 2 and Fig. 6 are the pending original mist image that has in the embodiment of the present invention; Fig. 5 is the result of a kind of image defogging method capable of non-use the present invention; From result contrast, can find out: the present invention can effectively process coloured image also can process gray level image, the brighter part of light can both clearly be shown, and there is no distortion.

Specific embodiment described herein is only to the explanation for example of the present invention's spirit.Those skilled in the art can make various modifications or supplement or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.

Claims (1)

1. the single image defogging method capable based on greasy weather imaging model, is characterized in that, comprises the following steps:

Step 1: read in the original mist image I (x) that has;

Step 2: image mist elimination based on greasy weather imaging model I (x), the coordinate of putting on x presentation video, I (x) represents the image of observing, t (x) is the transmissivity that a scalar represents light, A represents the color of atmosphere light, J (x) true picture recovering of indicating;

I (x)=J (x) t (x) R+ (1-t (x)) A formula one

R is decomposed along both direction, is respectively A direction and perpendicular to A direction; Residual vector after A Directional Decomposition is represented with R':

$I\left(x\right)=t\left(x\right)l^{\&prime;}\left(x\right)(\frac{R^{\&prime;}}{\left|\right|R^{\&prime;}\left|\right|}+\mathrm{\&eta;}\frac{A}{\left|\right|A\left|\right|})+(1-t\left(x\right))A$ Formula two

Wherein, r is a three-dimensional RGB vector, presentation surface reflection coefficient, and R' represents that R is along the component of A Directional Decomposition; L (x) is a scalar, represents how much catoptrically from body surface, l'(x) represents that l (x) is along the component of A Directional Decomposition;

Wherein, transmissivity t (x) is based on following formula:

$t\left(x\right)=1-\frac{(I_A\left(x\right)-\mathrm{\&eta;}{I}_{R^{\&prime;}}\left(x\right))}{\left|\right|A\left|\right|}$ Formula three

Wherein, η is based on following formula:

$\mathrm{\&eta;}=\frac{C_{\mathrm{\&Omega;}}(I_A,h)}{C_{\mathrm{\&Omega;}}(I_{R^{\&prime;}},h)}$ Formula four

Wherein c _Ω(I _a, h) and C _Ω(I _r', h) represent respectively I _awith h and I _r'covariance with h;

Step 3: obtain transmissivity t (x) value based on formula four and formula three, and t (x) is normalized:

$t^{\&prime;}=\frac{{t-t}_{\min }}{t_{\max }-t_{\min }}$ Formula five

And transmissivity t' and formula one based on after adjusting, according to formula, draw final J (x):

$J\left(x\right)=\frac{I\left(x\right)-\left({1-t}^{\&prime;}\left(x\right)\right)\&times;A}{t^{\&prime;}\left(x\right)}.$