CN105989583A - Image defogging method - Google Patents

Abstract

The invention discloses an image defogging method. The method uses a new transmittance estimation method to replace a soft matting method, the transmittance is solved through gain intervention on the three-primary color channel minimum component for a foggy image, on the premise of avoiding halation and fast effects, the approximate dark channel image intensity is solved quickly, the real-time performance of the algorithm is thus improved, a quad-tree subdivision method is used for estimating the atmospheric light value, the atmospheric light value precision is improved, and finally, through the transmittance and the atmospheric light value, defogging is carried out on multiple images. The method of the invention improves the real-time performance of defogging and the fidelity of a defogged image.

Description

A kind of image defogging method

Technical field

The invention belongs to Digital Image Processing and technical field of machine vision, particularly to a kind of image defogging method.

Background technology

Along with development and the progress of living standards of the people of science and technology, real life uses increasing vision System, such as: monitoring system, intelligent transportation system etc., human life quality and demand gradually step up, these application systems and the mankind Living closely related, direct image in human lives.

The reduction of air quality and reduce life and the industrial expansion to people of visibility cause the biggest shadow Ring.Absorbed by turbid media due to light and scatter, out of doors the picture of shooting under vile weather so that the intensity of light reduces, The light strong production causing optical pickocff to receive changes, and the fidelity ultimately resulting in color declines, and there is serious cross-color With skew, and the reduction of picture contrast, image detail information loses, and definition is inadequate, and the identification of pictorial information is significantly Reduce.Additionally, most automatic system, it is strongly dependent on the definition of input picture, and the image of degeneration can cause automatically System cisco unity malfunction, affects and limits the work such as highway vision monitoring, intelligent navigation, remote sensing monitoring.

Therefore, to have mist image mist elimination process have broad application prospects, can be applicable to underwater photograph technical, take photo by plane, family Outer monitoring even medical images etc., image and video through mist elimination process have more value, are conducive to many image understandings Applying (such as aerial image) with computer vision, image is classified, and image/video is retrieved, and remote sensing and video analysis and identification, to people Life bring a lot of facilities.In image processing field, the starting of image mist elimination is the most later.Chinese scholars carries at present The method of the image mist elimination gone out is not the most the most perfect, it would be highly desirable to improve, and specifically how improves the real-time of mist elimination process Property and fidelity be solve mist elimination problem key.

Summary of the invention

In order to solve the technical problem that above-mentioned background technology proposes, it is desirable to provide a kind of image defogging method, adopt By the absorbance improved and the computational methods of air light value, improve real-time and the precision of mist elimination image of mist elimination.

In order to realize above-mentioned technical purpose, the technical scheme is that

A kind of image defogging method, comprises the following steps:

(1) calculate and have the absorbance of mist image:

T (x, y)=(1-I^m(x,y))+gρ (1)

In formula (1), t (x, y) is absorbance, (x, y) represents pixel coordinate, and ρ is calibration factor, and g is gain constant:

$g=\frac{\underset{x,y}{\overset{\left|I^m\right|-1}{\Σ}}d(x,y)}{\left|d\right|}---\left(2\right)$

In formula (2), | I^m| and | d | is I respectively^mWith the sum of all pixels of d, corresponding I^mAnd d:

$I^m(x,y)=\underset{c\∈\{R,G,B\}}{min}{I}_c(x,y)---\left(3\right)$

D (x, y)=I^m(x,y)-I^d(x,y) (4)

In formula (3), I_cRepresent the Color Channel of mist elimination image；

In formula (4),Ω (x, y) represent with (x, y) centered by square Region, min represents and takes minima；

(2) use the air light value of quaternary tree close classification estimated mist image, specifically comprise the following steps that

A () will have mist image division to become some equal-sized rectangular sub blocks；

B () calculates the average pixel value of each rectangular sub blocks, retain the rectangular sub blocks that average pixel value is maximum, and by this square The average pixel value of shape sub-block is designated as max pixel value A_max；

If (c) max pixel value A_maxLess than pixel threshold a preset, and max pixel value A_maxCorresponding rectangular sub blocks Size more than or equal to the minimum window size preset, then returns to step (a), Further Division rectangular sub blocks, otherwise enters step (d)；

D image is transformed into YCbCr space from rgb space by (), and corresponding to the final rectangular sub blocks retained, select this square The maximum of shape sub-block luminance component is as air light value A_c；

(3) the air light value that the absorbance obtained according to step (1) and step (2) obtain, carries out mist elimination process to image:

$J(x,y)=\frac{I(x,y)-A_c}{max\left(t\right(x,y),t_0)}+A_c---\left(5\right)$

In formula (5), J represents the image after mist elimination process, and I is for having mist image, t₀For absorbance threshold value, max represents and takes Big value.

Further, in step (1), the span of calibration factor ρ is [0.8,1].

Further, the value of calibration factor ρ is 0.9.

Further, in step (3), absorbance threshold value t₀Span be [0.05,0.15].

Further, absorbance threshold value t₀Value be 0.1.

The beneficial effect that employing technique scheme is brought:

The present invention, compared in original dark algorithm, uses new absorbance estimation algorithm to replace soft stingy figure method, passes through There is the gain intervention of primary display channels minimum component of mist image to try to achieve absorbance, can be before avoiding halation and fast effect Put, quickly try to achieve approximation dark channel image intensity, thus improve the real-time of algorithm, and use the estimation of quaternary tree close classification big Gas light value so that the precision of air light value improves, thus reaches the effect of the image mist elimination that real-time is good and fidelity is high.

Accompanying drawing explanation

Fig. 1 is the basic flow sheet of the present invention.

Fig. 2 is the flow chart calculating atmosphere light value in the present invention.

Detailed description of the invention

Below with reference to accompanying drawing, technical scheme is described in detail.

As it is shown in figure 1, a kind of image defogging method, comprise the following steps:

Step 1, calculate and have the absorbance of mist image:

T (x, y)=(1-I^m(x,y))+gρ (1)

In formula (1), (x, y) is absorbance to t, and (x, y) represents pixel coordinate, and ρ is calibration factor, and the value of ρ generally exists In [0.8,1], the present embodiment can make ρ=0.9, and g is gain constant:

$g=\frac{\underset{x,y}{\overset{\left|I^m\right|-1}{\Σ}}d(x,y)}{\left|d\right|}---\left(2\right)$

In formula (2), | I^m| and | d | is I respectively^mWith the sum of all pixels of d, corresponding I^mAnd d:

$I^m(x,y)=\underset{c\∈\{R,G,B\}}{min}{I}_c(x,y)---\left(3\right)$

D (x, y)=I^m(x,y)-I^d(x,y) (4)

In formula (3), I_cRepresent the Color Channel of mist elimination image；

In formula (4),Ω (x, y) represent with (x, y) centered by square Region, min represents and takes minima.

Step 2, use quaternary tree close classification estimated mist image air light value, concrete steps as shown in Figure 2:

A () will have mist image division to become some equal-sized rectangular sub blocks；

B () calculates the average pixel value of each rectangular sub blocks, retain the rectangular sub blocks that average pixel value is maximum, and by this square The average pixel value of shape sub-block is designated as max pixel value A_max；

If (c) max pixel value A_maxLess than pixel threshold a preset, and max pixel value A_maxCorresponding rectangular sub blocks Size more than or equal to the minimum window size preset, then returns to step (a), Further Division rectangular sub blocks, otherwise enters step (d)；

D image is transformed into YCbCr space from rgb space by (), and corresponding to the final rectangular sub blocks retained, select this square The maximum of shape sub-block luminance component is as air light value A_c。

The air light value that step 3, the absorbance obtained according to step 1 and step 2 obtain, carries out mist elimination process to image:

$J(x,y)=\frac{I(x,y)-A_c}{max\left(t\right(x,y),t_0)}+A_c---\left(5\right)$

In formula (5), J represent mist elimination process after image, I is for there being mist image, and max represents and takes maximum, t₀For absorbance Threshold value, t₀Span be [0.05,0.15], the present embodiment can make t₀=0.1, be used for limiting absorbance t (x, y), due to As t, (x, when y) → 0, (x, y) (x, y) also can level off to 0 to t to J so that mist elimination image comprises noise, for avoiding sending out of this situation Raw, t is set₀Improve the quality of mist elimination image.

Above example is only the technological thought that the present invention is described, it is impossible to limit protection scope of the present invention with this, every The technological thought proposed according to the present invention, any change done on the basis of technical scheme, each fall within scope Within.

Claims (5)

1. an image defogging method, it is characterised in that comprise the following steps:

(1) calculate and have the absorbance of mist image:

T (x, y)=(1-I^m(x,y))+gρ (1)

In formula (1), t (x, y) is absorbance, (x, y) represents pixel coordinate, and ρ is calibration factor, and g is gain constant:

$g=\frac{\underset{x,y}{\overset{\left|I^m\right|-1}{\Σ}}d(x,y)}{\left|d\right|}---\left(2\right)$

In formula (2), | I^m| and | d | is I respectively^mWith the sum of all pixels of d, corresponding I^mAnd d:

$I^m(x,y)=\underset{c\∈\{R,G,B\}}{min}{I}_c(x,y)---\left(3\right)$

D (x, y)=I^m(x,y)-I^d(x,y) (4)

In formula (3), I_cRepresent the Color Channel of input picture；

In formula (4),Ω (x, y) represent with (x, y) centered by square region, Min represents and takes minima；

(2) use the air light value of quaternary tree close classification estimated mist image, specifically comprise the following steps that

A () will have mist image division to become some equal-sized rectangular sub blocks；

B () calculates the average pixel value of each rectangular sub blocks, retain the rectangular sub blocks that average pixel value is maximum, and by this rectangle The average pixel value of block is designated as max pixel value A_max；

If (c) max pixel value A_maxLess than pixel threshold a preset, and max pixel value A_maxThe size of corresponding rectangular sub blocks More than or equal to the minimum window size preset, then return to step (a), Further Division rectangular sub blocks, otherwise enter step (d)；

D image is transformed into YCbCr space from rgb space by (), and corresponding to the final rectangular sub blocks retained, select this rectangle The maximum of Block Brightness component is as air light value A_c；

(3) the air light value that the absorbance obtained according to step (1) and step (2) obtain, carries out mist elimination process to image:

$J(x,y)=\frac{I(x,y)-A_c}{max\left(t\right(x,y),t_0)}+A_c---\left(5\right)$

In formula (5), J represents the image after mist elimination process, and I is for having mist image, t₀For absorbance threshold value, max represents and takes maximum.

A kind of image defogging method, it is characterised in that: in step (1), calibration factor ρ takes Value scope is [0.8,1].

A kind of image defogging method, it is characterised in that: the value of calibration factor ρ is 0.9.

A kind of image defogging method, it is characterised in that: in step (3), absorbance threshold value t₀'s Span is [0.05,0.15].

A kind of image defogging method, it is characterised in that: absorbance threshold value t₀Value be 0.1.