CN105205794A - Synchronous enhancement de-noising method of low-illumination image - Google Patents

Abstract

The invention discloses a synchronous enhancement de-noising method of a low-illumination image. The synchronous enhancement de-noising method comprises the following steps: contrast enhancement and de-noising operations are simultaneously performed on the low-illumination image by using a noise-considered fog-degraded image model and with combination of an iterative associated bilateral filtering method; during the iterative initialization phase, the dark channel prior theory is used to obtain the global airglow, the transmissivity and the scene light initial estimated value of the reverse image of the low-illumination image; alternative modification is performed on parameters of the fog-degraded image by using the iterative associated bilateral filtering method, and detailed compensation is further performed on results obtained after several rounds of iteration by using a quotient image; when the first round of iteration is carried out by using the iterative associated bilateral filtering method, the oriented image is the noise image self, and during subsequent iterative processes, the result obtained after each round of iteration is taken as the oriented image of the next round of iteration; finally, re-reversing operation is performed on the iterative result to obtain the enhanced restored image. According to the synchronous enhancement de-noising method of the low-illumination image provided by the invention, the visualization effect of the low-illumination image is improved while the image noise is removed, and a good visual effect is further achieved.

Description

A kind of synchronous enhancing denoising method of low-light (level) image

Technical field

The present invention relates to image enhaucament, noise-removed technology field, specifically a kind of synchronous enhancing denoising method of low-light (level) image

Background technology

Under low-light (level) environment, image capture device obtain image not only can identification low, and containing much noise, the image deterioration that low-light (level) environment causes not only have impact on the identification of human eye to image, also make the computer vision system performances such as intelligent transportation, video monitoring and target identification be subject to larger impact, therefore low-light (level) image to be strengthened and noise reduction process has very important value; The gradation of image coverage collected due to low-light (level) environment is very narrow, and pixel value is in reduced levels, therefore the fundamental purpose strengthened low-light (level) image is the tonal range of expanded view picture, improve the overall brightness of image, so as the image information that originally cannot distinguish can by human eye or machine identify; Traditional image enchancing method can be divided into spatial domain Enhancement Method and the large class of frequency field Enhancement Method two: histogram equalization is classical spatial domain Enhancement Method, it effectively can strengthen picture contrast, but the method may cause brighter pixel supersaturation originally thus lose image structure information; Frequency field Enhancement Method, as wavelet transformation, after picture signal is transformed into frequency domain, processes the effect reaching enhancing to wavelet coefficient; Retinex method is a kind of image enchancing method based on retina cerebral cortex theory, image is divided into luminance graph and reflectogram two parts by it, the effect of enhancing is reached by reducing the impact of luminance graph on reflectogram, along with the proposition of Retinex theory, a large amount of researchers proposes relevant innovatory algorithm successively, as the multiple dimensioned Retinex etc. that single scale Retinex, multiple dimensioned Retinex, band color factor recover, these algorithms all can reach certain enhancing effect.

In recent years, proposition based on the enhancement method of low-illumination image of mist elimination technology opens the new way that low-light (level) strengthens, by the similarity of the reverse image and Misty Image that compare low-light (level) image, use dark primary defogging method capable to carry out process the low-light (level) frame of video of reversion and can obtain good visual effect, not only grey level is low also has the high feature of noise content for low-light (level) image, and most Enhancement Method noise while carrying out intensity-conversion also can amplify thereupon; There is a lot of research can complete enhancing and the denoising of image respectively at present, but for the enhancing of low-light (level) characteristics of image and denoising method little, enhancement method of low-illumination image based on dark primary mist elimination technology effectively can promote picture contrast, detailed information in outstanding image, but have the impact not considering noise due to dark primary mist elimination technology, cause strengthening rear noise and be significantly enlarged.

Application number is that the Chinese invention patent of CN201510260607 discloses a kind of video monitoring and the acquisition system that possess image enhancement functions, comprise the image acquiring device for obtaining image, analog-digital commutator AD, for strengthening to carrying out through analog-to-digital data image signal the image processing module processed, digiverter DA, video storage modules, monitor terminal, and for controlling the control axis module of whole system running status, the video monitoring of this invention and acquisition system adopt Embedded Hardware Platform, and software algorithm is optimized, thus the volume of image processing hardware equipment and power consumption are all greatly reduced, Real-time image enhancement can be realized, and convenient integrated with image capture device, modular design is also conducive to functional module and freely arranges in pairs or groups, more be applicable to application scenario, reduce purchase cost.

Application number is that the Chinese invention patent of CN201510329457 discloses a kind of grayscale image enhancement method based on retinal mechanisms, idiographic flow comprise estimate overall brightness determination algorithm auto-adaptive parameter, synthetic image brightness mapping graph, calculate brightness and strengthen image and Edge Enhancement process, first by the Luminance Distribution situation of overall dark areas, auto-adaptive parameter is estimated; Then respectively image is carried out to the brightness enhancing process of the overall situation, and drawn the modulation mapping graph of picture in its entirety by modulating function, calculate the result that brightness strengthens; The last Gaussian difference model based on adaptive scale realizes the enhancing at edge, and model dimension affected by contrast, and finally can strengthen more tiny texture information in bright areas, dark area then strengthens larger profile information.

Though above-mentioned invention disclosed can effectively be carried out contrast strengthen to image but all not consider the impact of noise conditions on image enhaucament, therefore, need to carry out creationary improvement to prior art.

Summary of the invention

The object of the present invention is to provide a kind of synchronous enhancing denoising method of low-light (level) image, utilize and consider that the operation of the contrast strengthen of low-light (level) image and noise-removal operation are carried out by the mist figure degradation model of noise and the associating bilateral filtering algorithm of iteration simultaneously, thus effectively promote picture contrast and restraint speckle, strengthen the visual effect of image.

For achieving the above object, the invention provides following technical scheme:

A kind of synchronous enhancing denoising method of low-light (level) image, enhancement algorithm for low-illumination image is the enhancing algorithm based on dark primary mist elimination technology, synchronous enhancing denoising operation is alternately revised mist figure degradation model parameter by the associating bilateral filtering of iteration and is completed, and comprises the following steps:

1) by original low-light (level) image I (x _i) input Computerized image processing system, and reversed, obtain the image I after reversing _inv(x _i);

2) according to dark primary priori theoretical, reverse image I is asked for _inv(x _i) in overall air light value A ^c;

3) according to reverse image I _inv(x _i) the initial transmission t of luminance graph computed image ⁰;

4) overall atmosphere light A step 2 tried to achieve ^cwith the initial transmission t that step 3 is tried to achieve ⁰substitute into the initial estimate that Misty Image degradation model obtains scene light

5) use the associating bilateral filtering method of iteration to replace the parameter revised in Misty Image degradation model, and use quotient image method to carry out details compensation to each result of taking turns.

6) by the scene light obtained final in step 5 reverse, obtain final enhancing denoising result.

As the further scheme of the present invention: in described step 1, to input low-light (level) image when carrying out reverse turn operation, reversion algorithm is as follows:

$I_{inv}^c\left(x_i\right)=255-I^c\left(x_i\right)$

Wherein, I represents the original low-light (level) image of input, I _invrepresent reverse image, a Color Channel in c representative image RGB tri-Color Channel.

As the further scheme of the present invention: described step 2 comprises the following steps:

A) to reverse image I _inv(x _i) each Color Channel do mini-value filtering, and each pixel is asked for the dark primary value of minimum value as this pixel of triple channel filter result, thus obtains the dark primary figure of reverse image;

B) 0.1% pixel that intensity level in all pixels in dark primary figure is maximum is chosen, by the position mark of these pixels out, at reverse image I _inv(x _i) position corresponding in three Color Channels, the intensity level of the point finding each passage the brightest is as the atmosphere light A of this Color Channel ^c.

As the further scheme of the present invention: in described step 3, ask for initial transmission t ⁰algorithm as follows:

t ⁰(x _i)＝C-Y(x _i)

In formula, C is the parameter for weakening luminance picture Y, and C span is [1.06,1.08], luminance picture Y (x _i) account form as follows:

Y(x _i)＝0.299×R+0.587×G+0.114×B

In formula: R, G, B be representative image RGB triple channel component value respectively.

As the further scheme of the present invention: in described step 4, the algorithm asking for scene light initial estimate is as follows

$J_{inv}^0\left(x_i\right)=\frac{I_{inv}\left(x_i\right)-A^c}{\max \left(t^0\right(x_i),\hat{t})}+A^c$

In formula, I _inv(x _i) be the reverse image of input picture, t ⁰(x _i) be the initial estimate of transmissivity, A ^cfor overall atmosphere light, for the lower limit of transmissivity, usually get 0.01.

As the further scheme of the present invention: the form of the Misty Image degradation model of the consideration noise used in described step 5 is as follows:

I _inv(x)＝J _inv(x)t(x)+A(1-t(x))+n(x)

As the further scheme of the present invention: the associating bilateral filtering method of the iteration in described step 5 is when first round iteration, and guiding figure is noise image self, in iterative process afterwards, each is taken turns the guiding figure of result as next round iteration of iteration.

As the further scheme of the present invention: comprise the following steps in described step 5:

A) initial value arranging transmissivity and scene light in iterative process is respectively t ⁰with juxtaposition iterations k=1;

B) revised scene light in last round of iteration is used revise the transmissivity t in epicycle iteration ^k, the algorithm revising transmissivity is as follows:

$t^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(J_{inv}^{k-1}\right(x_i)-A)}$

In formula, g _d(x _i-x) be spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

C) revised transmittance values t in epicycle iteration is used ^krevise the scene light in epicycle iteration the algorithm revising scene light is as follows:

$J_{inv}^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A(1-t^k\left(x_i\right)\left)\right)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)t^k\left(x_i\right)}$

In formula, g _d(x _i-x) spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

D) utilize quotient image to compensate to the detail section that each takes turns filter result, obtain the scene light correction result that epicycle iteration is final details backoff algorithm is as follows:

$J_{inv}^{kFinal}=(1-M)J_{inv}^{kDetail}{J}_{inv}^k+{\mathrm{MJ}}_{inv}^k$

In formula, M for balancing the weight shared by detail pictures, for quotient image, computing method are as follows:

$J_{inv}^{Detail}=\frac{J_{inv}^{k-1}+\mathrm{\ϵ}}{J_{inv}^k+\mathrm{\ϵ}}$

As the further scheme of the present invention: when revising transmissivity in described step 5, filter window chooses 15 × 15, when revising scene light, filter window chooses 7 × 7.

Compared with prior art, the invention has the beneficial effects as follows:

(1) the present invention proposes a kind of synchronous enhancing denoising method of low-light (level) image, utilize and consider that the operation of the contrast strengthen of low-light (level) image and noise-removal operation are carried out by the mist figure degradation model of noise and the associating bilateral filtering algorithm of iteration simultaneously, thus effectively promote picture contrast and restraint speckle, strengthen the visual effect of image;

(2) the present invention is often taking turns in the iterative process strengthening denoising, utilizes quotient image to compensate, make result comprise abundant detailed information to iteration result;

(3) the present invention takes into full account the unique characteristics of low-light (level) image, and the feature for low-light (level) image sets up physical model, synchronously completes enhancing, denoising operation.

Accompanying drawing explanation

Fig. 1 is the process flow diagram that low-light (level) image synchronization of the present invention strengthens denoise algorithm;

Fig. 2 is untreated photo.

Fig. 3 is that Fig. 2 adopts low-light (level) image synchronization of the present invention to strengthen the treatment effect figure of denoise algorithm.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the present invention, be clearly and completely described the technical scheme in the embodiment of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.

Refer to Fig. 1, a kind of synchronous enhancing denoising method of low-light (level) image, first use the overall atmosphere light of the mist elimination technology estimation low-light (level) image inversion image based on dark primary priori theoretical, the initial estimate of transmissivity and scene light; Then realize synchronously strengthening denoising by the associating bilateral filtering unknown parameter alternately revised in mist figure degradation model of iteration, last reverse image obtains final enhancing result, comprises the following steps:

1) by original low-light (level) image I (x _i) input Computerized image processing system, and reversed, obtain the image I after reversing _inv(x _i), to input low-light (level) image when carrying out reverse turn operation, reversion algorithm is as follows:

$I_{inv}^c\left(x_i\right)=255-I^c\left(x_i\right)$

2) according to dark primary priori theoretical, reverse image I is asked for _inv(x _i) in overall air light value A ^c, concrete steps are as follows:

A) to reverse image I _inv(x _i) each Color Channel do mini-value filtering, and each pixel is asked for the dark primary value of minimum value as this pixel of triple channel filter result, thus obtains the dark primary figure of reverse image;

B) 0.1% pixel that intensity level in all pixels in dark primary figure is maximum is chosen, by the position mark of these pixels out, at reverse image I _inv(x _i) position corresponding in three Color Channels, the intensity level of the point finding each passage the brightest is as the atmosphere light A of this Color Channel ^c;

3) according to reverse image I _inv(x _i) the initial transmission t of luminance graph computed image ⁰, ask for initial transmission t ⁰algorithm as follows:

t ⁰(x _i)＝C-Y(x _i)

In formula, C is the parameter for weakening luminance picture Y, and C span is [1.06,1.08], luminance picture Y (x _i) account form as follows:

Y(x _i)＝0.299×R+0.587×G+0.114×B

In formula: R, G, B be representative image RGB triple channel component value respectively.

4) overall atmosphere light A step 2 tried to achieve ^cwith the initial transmission t that step 3 is tried to achieve ⁰substitute into the initial estimate that Misty Image degradation model obtains scene light the algorithm asking for scene light initial estimate is as follows:

$J_{inv}^0\left(x_i\right)=\frac{I_{inv}\left(x_i\right)-A^c}{\max \left(t^0\right(x_i),\hat{t})}+A^c$

In formula, I _inv(x _i) be the reverse image of input picture, t ⁰(x _i) be the initial estimate of transmissivity, A ^cfor overall atmosphere light, for the lower limit of transmissivity, usually get 0.01;

5) use the associating bilateral filtering method of iteration to replace the parameter revised in the Misty Image degradation model considering noise, and use quotient image method to carry out details compensation to each result of taking turns, concrete steps are as follows:

A) initial value arranging transmissivity and scene light in iterative process is respectively t ⁰with juxtaposition iterations k=1;

B) revised scene light in last round of iteration is used revise the transmissivity t in epicycle iteration ^k, the algorithm revising transmissivity is as follows:

$t^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(J_{inv}^{k-1}\right(x_i)-A)}$

In formula, g _d(x _i-x) be spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

C) revised transmittance values t in epicycle iteration is used ^krevise the scene light in epicycle iteration the algorithm revising scene light is as follows:

$J_{inv}^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A(1-t^k\left(x_i\right)\left)\right)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)t^k\left(x_i\right)}$

In formula, g _d(x _i-x) spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

D) utilize quotient image to compensate to the detail section that each takes turns filter result, obtain the scene light correction result that epicycle iteration is final details backoff algorithm is as follows:

$J_{inv}^{kFinal}=(1-M)J_{inv}^{kDetail}{J}_{inv}^k+{\mathrm{MJ}}_{inv}^k$

In formula, M for balancing the weight shared by detail pictures, for quotient image, computing method are as follows:

$J_{inv}^{Detail}=\frac{J_{inv}^{k-1}+\mathrm{\ϵ}}{J_{inv}^k+\mathrm{\ϵ}}$

The associating bilateral filtering method of the iteration in described step 5 is when first round iteration, guiding figure is noise image self, in iterative process afterwards, each is taken turns the guiding figure of result as next round iteration of iteration, and filter window chooses 15 × 15 when revising transmissivity, during correction scene light, filter window chooses 7 × 7;

6) by the scene light obtained final in step 5 reverse, obtain final enhancing denoising result.

Embodiment:

The invention provides a kind of synchronous enhancing denoising method of low-light (level) image, while raising low-light (level) image viewing effect, effectively remove picture noise, effect of the present invention can be further illustrated by following experimental data:

Refer to Fig. 2-3, Fig. 2 is a width size is 1920 × 1080 low-light (level) images that with the addition of that standard deviation is the white Gaussian noise of 5, the net result of the synchronous enhancing denoising method process using the present invention to propose as shown in Figure 3, as can be seen from Fig. 2-3, the low-light (level) that the present invention proposes synchronously strengthens denoising method and effectively can promote integral image brightness and effectively remove picture noise, the sightless scene informations of script such as step not only can be observed out in enhancing result, and as the still high-visible impact not being subject to denoising operation of the detailed information such as the number-plate number.

The present invention proposes a kind of synchronous enhancing denoising method of low-light (level) image, utilize and consider that the operation of the contrast strengthen of low-light (level) image and noise-removal operation are carried out by the mist figure degradation model of noise and the associating bilateral filtering algorithm of iteration simultaneously, thus effectively promote picture contrast and restraint speckle, strengthen the visual effect of image; Often taking turns in the iterative process strengthening denoising, utilizing quotient image to compensate to iteration result, making result comprise abundant detailed information; Take into full account the unique characteristics of low-light (level) image, the feature for low-light (level) image sets up physical model, synchronously completes enhancing, denoising operation.

To those skilled in the art, obviously the invention is not restricted to the details of above-mentioned one exemplary embodiment, and when not deviating from spirit of the present invention or essential characteristic, the present invention can be realized in other specific forms.Therefore, no matter from which point, all should embodiment be regarded as exemplary, and be nonrestrictive, scope of the present invention is limited by claims instead of above-mentioned explanation, and all changes be therefore intended in the implication of the equivalency by dropping on claim and scope are included in the present invention.Any Reference numeral in claim should be considered as the claim involved by limiting.

In addition, be to be understood that, although this instructions is described according to embodiment, but not each embodiment only comprises an independently technical scheme, this narrating mode of instructions is only for clarity sake, those skilled in the art should by instructions integrally, and the technical scheme in each embodiment also through appropriately combined, can form other embodiments that it will be appreciated by those skilled in the art that.

Claims (9)

1. a synchronous enhancing denoising method for low-light (level) image, is characterized in that, comprise the following steps:

1) by original low-light (level) image I (x _i) input Computerized image processing system, and reversed, obtain the image I after reversing _inv(x _i);

2) according to dark primary priori theoretical, reverse image I is asked for _inv(x _i) in overall air light value A ^c;

3) according to reverse image I _inv(x _i) the initial transmission t of luminance graph computed image ⁰;

4) overall atmosphere light A step 2 tried to achieve ^cwith the initial transmission t that step 3 is tried to achieve ⁰substitute into the initial estimate that Misty Image degradation model obtains scene light

5) use the associating bilateral filtering method of iteration to replace the parameter revised in the Misty Image degradation model considering noise, and use quotient image method to carry out details compensation to each result of taking turns;

6) by the scene light obtained final in step 5 reverse, obtain final enhancing denoising result.

2. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, the initial estimate of described scene light ask for and carry out with noise remove simultaneously.

3. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, in described step 1, to input low-light (level) image when carrying out reverse turn operation, concrete reversion algorithm is as follows:

$I_{inv}^c\left(x_i\right)=255-I^c\left(x_i\right)$

Wherein, I represents the original low-light (level) image of input, I _invrepresent reverse image, a Color Channel in c representative image RGB tri-Color Channel.

4. the synchronous enhancing denoising method of low-light (level) image according to claim 1, it is characterized in that, described step 2 comprises the following steps enforcement:

A) to reverse image I _inv(x _i) each Color Channel do mini-value filtering, and each pixel is asked for the dark primary value of minimum value as this pixel of triple channel filter result, thus obtains the dark primary figure of reverse image;

B) 0.1% pixel that intensity level in all pixels in dark primary figure is maximum is chosen, by the position mark of these pixels out, at reverse image I _inv(x _i) position corresponding in three Color Channels, the intensity level of the point finding each passage the brightest is as the atmosphere light A of this Color Channel ^c.

5. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, in described step 3, ask for initial transmission t ⁰algorithm as follows:

t ⁰(x _i)＝C-Y(x _i)

In formula, C is the parameter for weakening luminance picture Y, and C span is [1.06,1.08], luminance picture Y (x _i) concrete account form as follows:

Y(x _i)＝0.299×R+0.587×G+0.114×B

In formula: R, G, B be representative image RGB triple channel component value respectively.

6. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, in described step 4, the algorithm asking for scene light initial estimate is as follows:

$J_{inv}^0\left(x_i\right)=\frac{I_{inv}\left(x_i\right)-A^c}{\max \left(t^0\right(x_i),\hat{t})}+A^c$

In formula, I _inv(x _i) be the reverse image of input picture, t ⁰(x _i) be the initial estimate of transmissivity, A ^cfor overall atmosphere light, for the lower limit of transmissivity, usually get 0.01.

7. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, the form of the Misty Image degradation model of the consideration noise used in described step 5 is as follows:

I _inv(x)＝J _inv(x)t(x)+A(1-t(x))+n(x)

The associating bilateral filtering method of the iteration in described step 5 is when first round iteration, and guiding figure is noise image self, in iterative process afterwards, each is taken turns the guiding figure of result as next round iteration of iteration.

8. the synchronous enhancing denoising method of low-light (level) image according to claim 1, it is characterized in that, described step 5 comprises the following steps enforcement:

A) initial value arranging transmissivity and scene light in iterative process is respectively t ⁰with juxtaposition iterations k=1;

B) revised scene light in last round of iteration is used revise the transmissivity t in epicycle iteration ^k, the algorithm revising transmissivity is as follows:

$t^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(t^{k-1}\right(x_i)-t^{k-1}(x\left)\right)\left(J_{inv}^{k-1}\right(x_i)-A)}$

In formula, g _d(x _i-x) be spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

C) revised transmittance values t in epicycle iteration is used ^krevise the scene light in epicycle iteration the algorithm revising scene light is as follows:

$J_{inv}^k\left(x\right)=\frac{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)\left(I_{inv}\right(x_i)-A(1-t^k\left(x_i\right)\left)\right)}{\underset{x_i\∈\mathrm{\Ω}\left(x\right)}{\Σ}{g}_d(x_i-x)g_r\left(J_{inv}^{k-1}\right(x_i)-J_{inv}^{k-1}(x\left)\right)t^k\left(x_i\right)}$

In formula, g _d(x _i-x) spatial domain kernel function, for codomain kernel function, Ω (x) is the neighborhood centered by x;

D) utilize quotient image to compensate to the detail section that each takes turns filter result, obtain the scene light correction result that epicycle iteration is final details backoff algorithm is as follows:

$J_{inv}^{kFinal}=(1-M)J_{inv}^{kDetail}{J}_{inv}^k+{\mathrm{MJ}}_{inv}^k$

In formula, M for balancing the weight shared by detail pictures, for quotient image, computing method are as follows:

$J_{inv}^{Detail}=\frac{J_{inv}^{k-1}+\mathrm{\ϵ}}{J_{inv}^k+\mathrm{\ϵ}}$

In formula, ε is used for the impact of attenuating noise.

9. the synchronous enhancing denoising method of low-light (level) image according to claim 1, is characterized in that, when revising transmissivity in described step 5, filter window size chooses 15 × 15, and when revising scene light, filter window size chooses 7 × 7.