CN102722864A - Image enhancement method - Google Patents

Abstract

The invention provides an image enhancement method, comprising steps that: a visible-light image and an infrared image are collected; reversible transformation is carried out on the brightness and the dimension of the visible-light image and the infrared image to obtain contrast information and texture information of the visible-light image and the infrared image; a mask is calculated on the basis of the saturation and the brightness of the visible-light image; the mask is used to migrate the visible-light image and the infrared image, and the contrast information for enhancing the visible-light image is calculated; the mask is used to migrate the visible-light image and the infrared image, and the texture information for enhancing the visible-light image is calculated; reversible transformation is carried out to obtain the brightness of the enhanced visible-light image; and the enhanced visible-light image is obtained on the basis of the brightness of the enhanced visible-light image, the saturation of the visible-light image and tone mixing. The image enhancement method of the embodiment of the invention can automatically realize high dynamic scene information recovery and achieve high qualified visible light image enhancement from two aspects of the contrast and the texture.

Description

A kind of image enchancing method

Technical field

The present invention relates to computer vision field, particularly a kind of image enchancing method.

Background technology

Because the restriction of visible light camera images acquired dynamic range, gather some zone of picture that the tangible scene of light and shade contrast possibly cause obtaining and cross bright and some zone is dark excessively.In order to obtain the image of HDR; Generally can gather the image of many different depth of exposures; Carry out the tone adjustment; Be about to the HDR luminance graph and be transformed into the figure of corresponding low-dynamic range with it, owing to need this image of image of different exposure time generally only to be applicable to static scene.Because original raw form contains more scene dynamics range information with respect to jpeg form commonly used; Also having a kind of method comparatively commonly used is exactly the artificial adjustment of image to the raw form that collects; It also is limited that but this method needs the dynamic range that a lot of manual works is participated in and the raw form collects, and can not collect the high multidate information of real scene fully.

Summary of the invention

The present invention is intended to one of solve the problems of the technologies described above at least.

For this reason, the objective of the invention is to propose a kind of image enchancing method that utilizes infrared image to recover real scene from contrast and texture two aspects.

Image enchancing method according to the embodiment of the invention comprises step: A. gathers visible images and infrared image, and wherein visible images aligns with the photocentre of infrared image; B. the brightness dimension of visible images is carried out reversible transformation, obtain visible images contrast information V _LWith visible images texture information V _D, and the brightness dimension of infrared image carried out reversible transformation, obtain infrared image contrast information N _LWith infrared image texture information N _DC. calculate masking-out W according to the saturation degree S of visible images and the brightness V of visible images, masking-out W is used to merge visible images and infrared image; D. according to visible images contrast information V _L, infrared image contrast information N _LWith masking-out W, calculate the contrast information V ' that strengthens visible images _LE. according to visible images texture information V _D, infrared image texture information N _DWith masking-out W, calculate the texture information V ' that strengthens visible images _DF. according to the contrast information V ' that strengthens visible images _LWith the texture information V ' that strengthens visible images _D, carry out reversible inverse transformation, the brightness V ' of the visible images that is enhanced; And G. mixes the visible images that is enhanced according to the brightness V ' that strengthens visible images and the saturation degree S of visible images and the tone H of visible images.

Method of the present invention is recovered the HDR of real scene through utilizing infrared image from contrast and two aspects of texture, realize that visible images strengthens, and this method can not need the artificial recovery of participating in realizing automatically the information of high dynamic scene.Advantage of the present invention is to utilize the method for calculating shooting in conjunction with infrared image, realizes that from contrast and texture two aspects high-quality visible images strengthens.

In one embodiment of the invention, reversible transformation is a wavelet transformation; Reversible contravariant is changed to inverse wavelet transform.

In one embodiment of the invention, in the visible images, saturation degree S crosses low and the too high or too low zone of brightness V, and masking-out W value is less.

In one embodiment of the invention; Step C comprises: C1. calculates initial masking-out

according to the brightness V of the saturation degree S of visible images and visible images and C2. is optimized calculating according to initial masking-out

, obtains masking-out W.

In one embodiment of the invention, calculate initial masking-out among the step C1

Formula be:

Wherein, W _s=e ^{-α | s-1|}, W _v=e ^{-β | v-0.5|}, α is that positive coefficient, β are positive coefficient.

In one embodiment of the invention, the method for calculating masking-out W is among the step C2:

$E\left(W\right)=W^T\mathrm{LW}+\mathrm{\λ}{(W-\stackrel{\‾}{W})}^T(W-\stackrel{\‾}{W})---\left(1\right)$

Wherein L is Laplce's matrix, and λ is the specification item coefficient, and (i, j) individual element is defined as in the matrix L

V wherein _iAnd V _jBe that visible images is at (i, brightness value j), δ _IjBe impulse function, in that (i is 1 j), and other point is 0, μ _kBe window w _kThe average of middle brightness of image value, ∑ _kBe window w _kMiddle brightness of image value variance, ε is the standardization parameter, | w _k| be window w _kThe number of middle element; Differentiate can obtain to equation (1)

$(L-\mathrm{\λU})W=\mathrm{\λ}\stackrel{\‾}{W}---\left(3\right)$

Then final masking-out W can obtain through finding the solution linear equation (3).

In one embodiment of the invention, strengthen the contrast information V ' of visible images among the step D _LComputing formula be: V ' _L=WV _L+ (1-W) V _{L '}, V wherein _{L '}Be new luminance graph information, new luminance graph information V _{L '}Gradient amplitude coupling through visible images and infrared image obtains.

In one embodiment of the invention, it is characterized in that new luminance graph information V _{L '}Computing method be: use

The gradient amplitude value of expression visible images,

The gradient amplitude value of expression infrared image, statistics V _GAnd N _GProbability histogram, utilize laplacian curve that histogram is carried out match, try to achieve the probability integral function of two curves, the probability integral function is mated, obtain the V ' of matching result _G, promptly obtain the gradient that new visible images brightness is tieed up

According to what obtain

With

Finding the solution Poisson equation finds the solution and obtains new luminance graph information V _{L '}

In one embodiment of the invention, strengthen the texture information V ' of visible images in the step e _DComputing formula be: V ' _D=WV _D+ (1-W) N _D

Aspect that the present invention adds and advantage part in the following description provide, and part will become obviously from the following description, or recognize through practice of the present invention.

Description of drawings

Above-mentioned and/or the additional aspect of the present invention and advantage from below in conjunction with accompanying drawing to becoming the description of embodiment obviously and understanding easily, wherein,

Fig. 1 is an image enchancing method process flow diagram according to an embodiment of the invention.

Embodiment

Describe embodiments of the invention below in detail, the example of said embodiment is shown in the drawings, and wherein identical from start to finish or similar label is represented identical or similar elements or the element with identical or similar functions.Be exemplary through the embodiment that is described with reference to the drawings below, only be used to explain the present invention, and can not be interpreted as limitation of the present invention.On the contrary, embodiments of the invention comprise and fall into appended spirit that adds the right claim and all changes, modification and the equivalent in the intension scope.

Image enchancing method of the present invention recovers the HDR of real scene through utilizing infrared image from contrast and two aspects of texture; Realize that former visible images strengthens, this method can not need the artificial recovery of participating in realizing automatically the information of high dynamic scene.Advantage of the present invention is to utilize the method for calculating shooting in conjunction with infrared image, realizes that from contrast and texture two aspects high-quality visible images strengthens.

Fig. 1 is an image enchancing method process flow diagram according to an embodiment of the invention.

As shown in Figure 1, image enchancing method comprises the steps:

Step S101 obtains the visible light and the infrared image that have alignd.

Particularly; Utilize hardware facility that the photocentre of visible light camera and infrared camera is alignd; Scene through spectroscope lets two cameras photograph is identical, or utilizes the more existing good camera that alignd to gather the visible images and the infrared image of same scene.

Step S102, the contrast and the texture information of extraction visible images and infrared image.

Particularly, the brightness dimension of visible images is carried out reversible transformation, obtain the visible images contrast information V of low-frequency range _LVisible images texture information V with high band _D, and the brightness dimension of infrared image carried out reversible transformation, obtain the infrared image contrast information N of low-frequency range _LInfrared image texture information N with high band _DIn a preferred embodiment of the invention, wavelet transformation is adopted in reversible transformation.

Step S103 calculates masking-out W according to the saturation degree S of visible images and the brightness V of visible images, and masking-out W is used to merge visible images and infrared image.

Particularly, the saturation degree S of image and brightness V are 0 to 1 value.In general, the zone of the zone of saturation degree S low excessively (being that the S value was near 0 o'clock) and brightness V too high (being that the V value is near 1) or low excessively (being that the V value is near 0) lacks texture information in the visible images, and contrast is gathered not enough.We make initial masking-out

Give less weights to these zones, can utilize

Obtain initial masking-out Value, wherein, W _s=e ^{-α | s-1|}, W _v=e ^{-β | v-0.5|}, wherein α and β are positive coefficient.Need to prove; The function that calculates initial masking-out

is also not exclusive; Other function also can use as long as satisfy the condition of " weights of the initial masking-out

than the zonule are given in place and the too high or too low place of brightness V to saturation degree S is low excessively in the visible images ".

After the value that obtains initial masking-out

; We obtain high-quality masking-out W by further optimization, optimize equation and are:

$E\left(W\right)=W^T\mathrm{LW}+\mathrm{\λ}{(W-\stackrel{\‾}{W})}^T(W-\stackrel{\‾}{W})---\left(1\right)$

Wherein L is Laplce's matrix, and λ is the specification item parameter.(i, j) individual element is defined as in the matrix L

$\underset{k|(i,j)\∈w_k}{\mathrm{\Σ}}({\mathrm{\δ}}_{\mathrm{ij}}-\frac{1}{\left|w_k\right|}(1+{(V_i-{\mathrm{\μ}}_k)}^T({\∑}_k+\frac{\mathrm{\ϵ}}{\left|w_k\right|})(V_j-{\mathrm{\μ}}_k)))---\left(2\right)$

V wherein _iAnd V _jBe that image is at (i, brightness value j), δ _IjBe impulse function, in that (i is 1 j), and other point is 0, μ _kAnd ∑ _kBe respectively window w _kThe average and the variance of middle brightness of image value, ε is the specification item coefficient, | w _k| be window w _kThe number of middle element.

Differentiate can obtain to equation (1)

$(L-\mathrm{\λU})W=\mathrm{\λ}\stackrel{\‾}{W}---\left(3\right)$

Wherein, the U representation unit matrix in the formula (3), the high-quality masking-out W after then optimizing can obtain through finding the solution linear equation (3).

Step S104 is according to visible images contrast information V _L, infrared image contrast information N _LWith masking-out W, calculate the contrast information V ' that strengthens visible images _LSpecifically comprise:

Experimental verification shows, the gradient of natural image and near-infrared image all meet laplacian distribution.Therefore, the present invention is mated the contrast of adjusting visible images with regard to the gradient amplitude of utilizing this two width of cloth image, wherein

The gradient amplitude of expression visible images,

The gradient amplitude value of expression infrared image, statistics V _GAnd N _GProbability histogram, utilize laplacian curve that histogram is carried out match, try to achieve the probability integral function of two curves, the probability integral function is mated, obtain the V ' of matching result _G, can obtain so new can be with the gradient of the brightness dimension of light image

According to what obtain

With Finding the solution Poisson equation finds the solution and obtains new visible images luminance graph V _{L '}The masking-out W that integrating step S103 obtains then tries to achieve the contrast information V ' of the final enhancing visible images after the migration _LFor

V′ _L=W·V _L+(1-W)·V _L′ （4）

Step S105 is according to visible images texture information V _D, infrared image texture information N _DWith masking-out W, calculate the texture information V ' that strengthens visible images _D

The masking-out W that utilizes step S103 to obtain, the texture information of fusion visible images and near-infrared image is through the texture migration of near-infrared image, the texture information V ' of the enhancing visible images that finally obtains _DThe computing formula of migration step is: V ' _D=WV _D+ (1-W) N _D

Step S106 merges the contrast information V ' that strengthens visible images _LWith texture information V ' _D, obtain the monochrome information V ' of high-quality enhancing visible images.

Particularly, the enhancing visible images contrast information V ' to moving through the contrast masking-out _LThe texture information V ' of the enhancing visible images that moved with the texture masking-out _D, the opposite variation of the reversible transformation that employing step S102 uses is to V ' _LAnd V ' _DCarrying out reversible inverse transformation obtains through strengthening the luminance graph V ' of visible images.In a preferred embodiment of the invention, reversible contravariant is changed to inverse wavelet transform.

Step S107 merges the monochrome information V ' that strengthens visible images and saturation degree S and the tone H of original visible images, obtains the final visible images that passes through after the infrared image enhancing.

In the description of this instructions, the description of reference term " embodiment ", " some embodiment ", " example ", " concrete example " or " some examples " etc. means the concrete characteristic, structure, material or the characteristics that combine this embodiment or example to describe and is contained at least one embodiment of the present invention or the example.In this manual, the schematic statement to above-mentioned term not necessarily refers to identical embodiment or example.And concrete characteristic, structure, material or the characteristics of description can combine with suitable manner in any one or more embodiment or example.

Although illustrated and described embodiments of the invention; For those of ordinary skill in the art; Be appreciated that under the situation that does not break away from principle of the present invention and spirit and can carry out multiple variation, modification, replacement and modification that scope of the present invention is accompanying claims and be equal to and limit to these embodiment.

Claims (9)

1. an image enchancing method is characterized in that, comprises step:

A. gather visible images and infrared image, wherein said visible images aligns with the photocentre of said infrared image;

B. the brightness dimension of said visible images is carried out reversible transformation, obtain visible images contrast information V _LWith visible images texture information V _D, and the brightness dimension of said infrared image carried out reversible transformation, obtain infrared image contrast information N _LWith infrared image texture information N _D

C. calculate masking-out W according to the saturation degree S of said visible images and the brightness V of said visible images, said masking-out W is used to merge said visible images and said infrared image;

D. according to said visible images contrast information V _L, said infrared image contrast information N _LWith said masking-out W, calculate the contrast information V ' that strengthens visible images _L

E. according to said visible images texture information V _D, said infrared image texture information N _DWith said masking-out W, calculate the texture information V ' that strengthens visible images _D

F. according to the contrast information V ' of said enhancing visible images _LTexture information V ' with said enhancing visible images _D, carry out reversible inverse transformation, the brightness V ' of the visible images that is enhanced; And

G. according to brightness V ' and the saturation degree S of said visible images and the tone H of said visible images of said enhancing visible images, mix the visible images that is enhanced.

2. image enchancing method as claimed in claim 1 is characterized in that, said reversible transformation is a wavelet transformation; Said reversible contravariant is changed to inverse wavelet transform.

3. image enchancing method as claimed in claim 1 is characterized in that, in the said visible images, said saturation degree S crosses low and the too high or too low zone of said brightness V, and said masking-out W value is less.

4. image enchancing method as claimed in claim 3 is characterized in that, said step C comprises:

C1. according to the brightness V of the saturation degree S of said visible images and said visible images calculate initial masking-out

and

C2. be optimized calculating according to said initial masking-out

, obtain masking-out W.

5. image enchancing method as claimed in claim 4 is characterized in that, calculates initial masking-out among the step C1

Formula be:

Wherein, W _s=e ^{-α | s-1|}, W _v=e ^{-β | v-0.5|}, α is that positive coefficient, β are positive coefficient.

6. image enchancing method as claimed in claim 4 is characterized in that, the computing formula of masking-out W is among the step C2:

$E\left(W\right)=W^T\mathrm{LW}+\mathrm{\λ}{(W-\stackrel{\‾}{W})}^T(W-\stackrel{\‾}{W})$

Wherein L is Laplce's matrix, and λ is the specification item coefficient, and (i, j) individual element is defined as in the matrix L

$\underset{k|(i,j)\∈w_k}{\mathrm{\Σ}}({\mathrm{\δ}}_{\mathrm{ij}}-\frac{1}{\left|w_k\right|}(1+{(V_i-{\mathrm{\μ}}_k)}^T({\∑}_k+\frac{\mathrm{\ϵ}}{\left|w_k\right|})(V_j-{\mathrm{\μ}}_k)))$

V wherein _iAnd V _jBe that said visible images is at (i, brightness value j), δ _IjBe impulse function, in that (i is 1 j), and other point is 0, μ _kBe window w _kThe average of middle brightness of image value, ∑ _kBe window w _kMiddle brightness of image value variance, ε is the standardization parameter, | w _k| be window w _kThe number of middle element;

Computing formula differentiate to said masking-out W can obtain linear equation

$(L-\mathrm{\λU})W=\mathrm{\λ}\stackrel{\‾}{W}$

Wherein, U is a unit matrix, and final said masking-out W obtains through finding the solution said linear equation.

7. image enchancing method as claimed in claim 1 is characterized in that, strengthens the contrast information V ' of visible images described in the said step D _LComputing formula be: V ' _L=WV _L+ (1-W) V _{L '}, V wherein _{L '}Be new luminance graph information, said new luminance graph information V _{L '}Gradient amplitude coupling through said visible images and said infrared image obtains.

8. image enchancing method as claimed in claim 7 is characterized in that, said new luminance graph information V _{L '}Computing method be:

Use The gradient amplitude value of representing said visible images,

The gradient amplitude value of representing said infrared image, statistics V _GAnd N _GProbability histogram, utilize laplacian curve that histogram is carried out match, try to achieve the probability integral function of two curves, the probability integral function is mated, obtain the V ' of matching result _G, promptly obtain the gradient that new visible images brightness is tieed up

According to what obtain

With

Finding the solution Poisson equation finds the solution and obtains said new luminance graph information V _{L '}

9. image enchancing method as claimed in claim 1 is characterized in that, strengthens the texture information V ' of visible images described in the said step e _DComputing formula be: V ' _D=WV _D+ (1-W) N _D

Images