CN113160066A - Low-illumination image efficient enhancement method - Google Patents

Low-illumination image efficient enhancement method Download PDF

Info

Publication number
CN113160066A
CN113160066A CN202110029801.0A CN202110029801A CN113160066A CN 113160066 A CN113160066 A CN 113160066A CN 202110029801 A CN202110029801 A CN 202110029801A CN 113160066 A CN113160066 A CN 113160066A
Authority
CN
China
Prior art keywords
image
low
illumination
images
enhancement method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Withdrawn
Application number
CN202110029801.0A
Other languages
Chinese (zh)
Inventor
李昌利
王阳阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hohai University HHU
Original Assignee
Hohai University HHU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hohai University HHU filed Critical Hohai University HHU
Priority to CN202110029801.0A priority Critical patent/CN113160066A/en
Publication of CN113160066A publication Critical patent/CN113160066A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/70Denoising; Smoothing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00Image enhancement or restoration
    • G06T5/73Deblurring; Sharpening
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10004Still image; Photographic image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10016Video; Image sequence
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10024Color image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20024Filtering details

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a high-efficiency enhancement method of a low-illumination image, which comprises the steps of firstly, adjusting each pixel value of the image by using two designed different transformation functions to obtain two enhanced images; simply fusing the two enhanced images to obtain an enhanced image; denoising the enhanced image by using a guide filter; and then sharpening the denoised image to enhance the details of the image. The method of the invention can improve the brightness and enhance the contrast of the image, and the algorithm is simple and has small calculation amount, thereby being beneficial to the real-time processing of the image. Compared with the traditional contrast enhancement algorithm, the method can enhance the brightness and the contrast of the low-illumination image without excessive enhancement, and can make the detail part of the image more obvious, so that the image has good visual effect.

Description

Low-illumination image efficient enhancement method
Technical Field
The invention relates to a low-illumination image efficient enhancement method, and belongs to the technical field of image processing.
Background
Generally, images or videos are acquired under the condition of sufficient light, but in practical use situations, images or videos do not always have to be acquired under low-illumination environments, for example, under low-illumination conditions such as indoor and night, due to insufficient illumination of an unnatural light source, reflected light of a target surface is weak, and light entering an imaging sensor is insufficient. The image of shooing under the low light level environment receives ambient light, the influence of equipment of shooing, and the photo visibility of shooing is relatively poor, and image identifiability is very low, is difficult to distinguish the detail in the image, influences visual effect, still influences subsequent further treatment moreover. Therefore, there is a need for brightness and contrast enhancement of low-illumination images to provide better visibility of the processed images.
Conventional methods for enhancing image brightness and contrast include methods based on linear stretching, logarithmic transformation, and gamma transformation. The gray level set of the image can be linearly stretched to 0-255 by adopting a linear stretching method, and the linear stretching can also stretch the brightness area of the image while enhancing the contrast of the image, so that the image is partially overexposed; by adopting the methods of logarithmic transformation and gamma transformation, the pixel values of dark areas of the image can be expanded, but the pixel values of bright areas of the image can be compressed, so that the details of the bright areas of the image are lost.
Guided filtering was first proposed by He et al at 2010 computer vision conference, an edge-preserving filter based on a local linear model. Fundamental principles of guided filtering: the output image is obtained by converting the input image and the guide image respectively in different linear relations, has a structural organization similar to the whole input image, and retains the edge information and the texture characteristics of the guide image. The guiding filtering is mainly applied to image smooth noise reduction, image defogging and the like. The literature applies the guiding filtering to the low-illumination image enhancement algorithm, and a good effect is achieved.
Disclosure of Invention
In order to solve the problems, the invention provides a low-illumination image efficient enhancement method.
The invention adopts the following technical scheme for solving the technical problems:
a low-illumination image efficient enhancement method comprises the following specific steps:
s1, carrying out normalization processing on the low-illumination image to obtain an image I;
s2, respectively adopting the following two transformation functions to transform the image I to obtain two enhanced images I1、I2Wherein the two transformation functions are:
Figure BDA0002891674080000011
wherein, I (x, y), I1(x, y) and I2(x, y) are images I, I, respectively1And I2The pixel value of the middle coordinate (x, y), gamma is an index of gamma conversion;
s3, two enhanced images I1、I2Fusing to obtain an image out;
s4, denoising the image out by using a guide filter to obtain an image q1
S5, for image q1Sharpening is carried out, and the sharpening result is the final output image Iout
Further, the normalization in S1 is specifically to normalize the pixel values of the pixels in the low-illuminance image from [0, 255] to within a [0, 1] interval.
Further, the fusion formula in S3 is out (x, y) ═ 1-a) I1(x,y)+aI2(x, y), where out (x, y) is the pixel value of the coordinate (x, y) in the image out, a is a weighting coefficient, a is mean { I }, and mean { I } represents the pixel mean value of the image I.
Further, S3 includes inverse normalizing the fused image.
Further, the inverse normalization specifically includes: and multiplying the pixel value of the pixel point in the fused image by 255 and rounding down.
Further, the input image is image out in S4, and the guide image is image I.
Further, the formula for sharpening in S5 is: i isout=λ(q1-q2)+q2Wherein the parameter lambda > 1, q2For using a guided filter on an image q1And (5) denoising a result.
Further, the guide image and the input image are both images q1
Compared with the prior art, the invention adopting the technical scheme has the following technical effects: the invention uses two different transformation functions to perform low-illumination image fusion enhancement. Then denoising the enhanced image by using a guide filter; and finally sharpening the denoised image. The method of the invention can improve the brightness and enhance the contrast of the image, and the algorithm is simple and has small calculation amount, thereby being beneficial to the real-time processing of the image. Compared with the traditional contrast enhancement algorithm, the method can enhance the brightness and the contrast of the low-illumination image without excessive enhancement, and can make the detail part of the image more obvious, so that the image has good visual effect.
Drawings
FIG. 1 is a flow chart of a method for enhancing low-illumination image efficiency according to the present invention;
fig. 2(a) is a low illumination image;
FIG. 2(b) is the image of FIG. 2(a) after undergoing the PIE algorithm;
FIG. 2(c) is the image of FIG. 2(a) after the NPE _ TIP algorithm;
FIG. 2(d) is the image of FIG. 2(a) obtained by the algorithm of the present invention;
FIG. 3(a) is an old photograph;
FIG. 3(b) is the image of FIG. 3(a) after the PIE algorithm;
FIG. 3(c) is the image of FIG. 3(a) after the NPE _ TIP algorithm;
FIG. 3(d) is the image of FIG. 3(a) obtained by the algorithm of the present invention.
Detailed Description
The invention and its advantageous effects are explained in detail below with reference to the drawings and the respective embodiments.
As shown in fig. 1, the low-illuminance image enhancement method of the present invention includes:
firstly, acquiring a low-illumination image, and performing normalization operation on all pixel values to obtain an image I.
The low-illumination image can be shot by a digital camera in an environment with insufficient illumination to obtain an original low-illumination image, and the image obtained by the digital camera is generally a digital image. The digital camera can be installed on a digital camera, an intelligent mobile phone, a tablet computer, various monitoring cameras and the like. Besides taking images by a digital camera to obtain original digital images, digital images can be obtained from others by copying or network transmission.
The normalization process here can be implemented using known techniques in order to normalize all pixel values of the low-illuminance image from [0, 255] to within the [0, 1] interval. All pixel values of the obtained low-illuminance image are normalized, for example, using an im2double () function, and the pixel values can be normalized from [0, 255] to within the [0, 1] interval. Other well-known methods for normalizing the pixel values from 0, 255 to 0, 1 intervals may be used, and are not illustrated here.
And secondly, processing each normalized pixel value by using two different designed transformation functions respectively, thereby obtaining two enhanced images.
Two different transformation functions are shown in equation 1 below:
Figure BDA0002891674080000031
where (x, y) is the specific location of the pixel, I (x, y) is the pixel value before transformation1(x,y)、I2(x, y) are the pixel values after the transform function 1 and transform function 2 processing, respectively. γ is an exponent of the gamma transformation, similar to the conventional gamma transformation in the present invention we set γ to 0.4545;
by the above conversion processing, the dark pixel value can be increased and the bright pixel value can be decreased, while the pixel values of 0 and 1 are kept unchanged, so that the pixel values between (0, 1) are subjected to nonlinear conversion, and the brightness and the contrast of the image can be improved. In the invention, the pixel value after normalization is in the interval of [0, 1], so that the pixel value is compared with 0.5, the pixel value with the value less than 0.5 is a dark pixel value, and the pixel value with the value more than 0.5 is a bright pixel value.
And thirdly, carrying out fusion processing and inverse normalization on the two enhanced images to obtain an enhanced image out.
The fusion mode is as follows:
out(x,y)=(1-a)I1(x,y)+aI2(x,y) (2)
where out (x, y) represents a pixel value after the fusion process, where a ═ mean { I } represents the pixel mean value of the image I.
Based on formula 2, all pixel values on the two enhanced images are traversed to perform calculation, so that all pixel values on the fused image obtained after fusion processing are obtained.
The method for performing inverse normalization on all pixel values on the fusion image obtained by the fusion processing comprises the following steps: each pixel value on the fused image is multiplied by 255 and rounded down, thus resulting in a new enhanced image.
The rounding-down operation may be performed by using a well-known rounding-down function, which refers to taking an integer smaller than the calculation result when the calculation result is not an integer, and directly taking an integer when the calculation result is an integer.
And fourthly, denoising the image out by using the guide filtering.
The input image is image out and the guide image is image I, taking a filter radius of 8 and an epsilon of 0.0004. Suppose an output image q1And a window omega with radius r and taking a pixel point k as a center with the guide image IrThe following linear transformation relationship exists:
Figure BDA0002891674080000041
wherein I represents the ith pixel point, IiThe pixel value of the ith pixel point in the representative image I, and the linear coefficient (a)k,bk) Is a constant, q1I.e. the de-noised image, q1iRepresentative image q1The pixel value of the ith pixel point;
linear coefficient (a)k,bk) The gradient is calculated for equation 3:
Figure BDA0002891674080000042
that is, the edge information of the output image and the guide image has a linear relationship, and a linear coefficient is obtained by minimizing a cost function, wherein the cost function is defined as:
Figure BDA0002891674080000043
the linear coefficient (a) is obtained from equation 4k,bk) In order to realize the purpose,
Figure BDA0002891674080000044
bk=μout,r(k)-akμI,r(k) (6)
where μ is the mean value, representing the matrix multiplication. Mu.sI·out,r(k) Is the mean, μ, of the image I multiplied by the image out within Ω rout,r(k) Is at omegarThe mean value of the intra-picture I,
Figure BDA0002891674080000045
is at omegarVariance of intra image I.
Due to the linear coefficients (a) solved in each window containing pixel ik,bk) Different values, so that the windows are solved firstk,bk) Average value of the values, and then the result is used as (a) of the pixel point ik,bk) The value, therefore equation 3 is redefined as:
Figure BDA0002891674080000046
wherein | Ωr(i) L is window omegar(i) Number of pixels (2)
The fifth step, to the image q1Sharpening is performed in order to enhance the details of the image and increase the contrast.
Computing the base layer with a guided filter, the detail layer being the image q1The base layer is subtracted.
Specifically, the guide image and the input image are both the image q when the base layer is calculated1The formula used for guided filtering is the same as the above formula for guided filtering, with the result that q is obtained2
Specifically, the formula for sharpening is:
Iout=λ(q1-q2)+q2,λ>1 (8)
specifically, the best results are obtained when λ ∈ [2, 3 ].
Image enhancement time versus quality assessment:
in order to verify that the speed of the method of the invention for low-illumination images is faster than that of the PIE algorithm and the NPE _ TIP algorithm, the experiment searches 500 pictures from the network, processes the 500 pictures by the three algorithms respectively, calculates the running time of each algorithm, and compares the average value of the 500 pictures processed by each algorithm conveniently, as shown in Table 1.
TABLE 1 average run time of the three algorithms
PIE NPE_TIP The invention
Mean value (seconds) 1.064405 11.09669 0.87303
In order to verify that the enhancement effect of the method of the invention on the low-illumination image is better than that of the PIE algorithm and the NPE _ TIP algorithm, the three enhancement algorithms are used for respectively enhancing the two low-illumination images, and the enhanced images are compared.
The effect contrast after the enhancement of several algorithms is given, the brightness and the color of the low-illumination image shown in fig. 2(a) are greatly improved after the enhancement of the several algorithms, and the visual effect is better. As shown in fig. 2(b), the brightness of the PIE enhanced is improved, but not enough; as shown in fig. 2(c), the image brightness enhanced by the NPE _ TIP algorithm is improved, and the part of the column in the picture is brighter than the image processed by the PIE, but this method also brings noise; as shown in fig. 2(d), the image enhanced by the algorithm of the present invention is better than PIE and NPE _ TIP in brightness, color, and detail.
The old photo shown in fig. 3(a) is enhanced in quality by PIE as shown in fig. 3(b), but the image is still dark in part; the image after enhancement by the NPE _ TIP algorithm is improved in brightness compared to PIE as shown in fig. 3 (c); the overall brightness after the enhancement by the algorithm of the present invention is better and more detailed than the two algorithms as shown in fig. 3 (d).
In conclusion, the image quality of the enhanced algorithm provided by the invention is better than that of PIE and NPE _ TIP.
Finally, it should be pointed out that: the above examples are only for illustrating the technical solutions of the present invention, and are not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (8)

1.一种低照度图像高效增强方法,其特征在于,具体步骤如下:1. a low-illumination image high-efficiency enhancement method, is characterized in that, concrete steps are as follows: S1,对低照度图像进行归一化处理,得到图像I;S1, normalize the low-illumination image to obtain image I; S2,对图像I分别采用以下两种变换函数进行变换,得到两个增强图像I1、I2,其中,两种变换函数为:S2, the following two transformation functions are used to transform the image I respectively to obtain two enhanced images I 1 and I 2 , wherein the two transformation functions are:
Figure FDA0002891674070000011
Figure FDA0002891674070000011
其中,I(x,y)、I1(x,y)和I2(x,y)分别为图像I、I1和I2中坐标(x,y)的像素值,γ为伽马变换的指数;Among them, I(x, y), I 1 (x, y) and I 2 (x, y) are the pixel values of the coordinates (x, y) in the images I, I 1 and I 2 , respectively, and γ is the gamma transformation index; S3,将两个增强图像I1、I2进行融合,得到图像out;S3, fuse the two enhanced images I 1 and I 2 to obtain an image out; S4,使用导向滤波器对图像out进行去噪,得到图像q1S4, use the guided filter to denoise the image out to obtain the image q 1 ; S5,对图像q1进行锐化,锐化的结果即为最终的输出图像IoutS5 , sharpen the image q 1 , and the sharpening result is the final output image I out .
2.如权利要求1所述的一种低照度图像高效增强方法,其特征在于,S1中的归一化具体为将低照度图像中像素点的像素值从[0,255]归一化到[0,1]区间内。2. The high-efficiency enhancement method for low-illumination images according to claim 1, wherein the normalization in S1 is specifically normalizing the pixel values of pixels in the low-illumination image from [0, 255] to in the interval [0, 1]. 3.如权利要求1所述的一种低照度图像高效增强方法,其特征在于,S3中的融合公式为out(x,y)=(1-a)I1(x,y)+aI2(x,y),其中out(x,y)为图像out中坐标(x,y)的像素值,a为加权系数,a=mean{I},mean{I}表示图像I的像素均值。3. The method for efficiently enhancing low-illuminance images according to claim 1, wherein the fusion formula in S3 is out(x, y)=(1-a)I 1 (x, y)+aI 2 (x, y), where out(x, y) is the pixel value of the coordinate (x, y) in the image out, a is the weighting coefficient, a=mean{I}, mean{I} represents the pixel mean of the image I. 4.如权利要求1所述的一种低照度图像高效增强方法,其特征在于,S3中还包括将融合后的图像进行逆归一化。4 . The high-efficiency enhancement method for low-illumination images according to claim 1 , wherein S3 further comprises performing inverse normalization on the fused image. 5 . 5.如权利要求4所述的一种低照度图像高效增强方法,其特征在于,逆归一化具体为:将融合后的图像中像素点的像素值乘以255并向下取整。5 . The high-efficiency enhancement method for a low-illumination image according to claim 4 , wherein the inverse normalization is specifically: multiplying the pixel value of the pixel point in the fused image by 255 and rounding it down. 6 . 6.如权利要求1所述的一种低照度图像高效增强方法,其特征在于,S4中输入图像为图像out,导向图像为图像I。6. The high-efficiency enhancement method for low-illuminance images as claimed in claim 1, wherein the input image in S4 is image out, and the guide image is image I. 7.如权利要求1所述的一种低照度图像高效增强方法,其特征在于,S5中锐化的公式为:Iout=λ(q1-q2)+q2,其中,参数λ>1,q2为使用导向滤波器对图像q1进行去噪的结果。7. The high-efficiency enhancement method for low-illumination images according to claim 1, wherein the formula for sharpening in S5 is: I out =λ(q 1 -q 2 )+q 2 , wherein the parameter λ > 1, q 2 are the results of denoising the image q 1 using a guided filter. 8.如权利要求7所述的一种低照度图像高效增强方法,其特征在于,导向图像和输入图像均为图像q18 . The high-efficiency enhancement method for low-illumination images according to claim 7 , wherein the guide image and the input image are both images q 1 .
CN202110029801.0A 2021-01-11 2021-01-11 Low-illumination image efficient enhancement method Withdrawn CN113160066A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110029801.0A CN113160066A (en) 2021-01-11 2021-01-11 Low-illumination image efficient enhancement method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110029801.0A CN113160066A (en) 2021-01-11 2021-01-11 Low-illumination image efficient enhancement method

Publications (1)

Publication Number Publication Date
CN113160066A true CN113160066A (en) 2021-07-23

Family

ID=76878304

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110029801.0A Withdrawn CN113160066A (en) 2021-01-11 2021-01-11 Low-illumination image efficient enhancement method

Country Status (1)

Country Link
CN (1) CN113160066A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114298916A (en) * 2021-11-11 2022-04-08 电子科技大学 X-Ray image enhancement method based on gray stretching and local enhancement
CN118658194A (en) * 2024-08-20 2024-09-17 成都贝尔通讯实业有限公司 A smart community management method, system and program product

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114298916A (en) * 2021-11-11 2022-04-08 电子科技大学 X-Ray image enhancement method based on gray stretching and local enhancement
CN118658194A (en) * 2024-08-20 2024-09-17 成都贝尔通讯实业有限公司 A smart community management method, system and program product
CN118658194B (en) * 2024-08-20 2024-10-29 成都贝尔通讯实业有限公司 Smart community management method, system and program product

Similar Documents

Publication Publication Date Title
US11127122B2 (en) Image enhancement method and system
CN111292257B (en) A Retinex-based Image Enhancement Method in Dark Vision Environment
CN111986120A (en) Low-illumination image enhancement optimization method based on frame accumulation and multi-scale Retinex
Kapoor et al. Fog removal in images using improved dark channel prior and contrast limited adaptive histogram equalization
CN104240194A (en) Low-light-level image enhancement algorithm based on parabolic function
CN110047055B (en) An infrared image detail enhancement and denoising method
CN116823686B (en) Night infrared and visible light image fusion method based on image enhancement
CN115578297A (en) A Generalized Attenuation Image Enhancement Method with Adaptive Color Compensation and Detail Optimization
CN112116536A (en) Low-illumination image enhancement method and system
CN108564544A (en) Image Blind deblurring based on edge perception combines sparse optimization method
CN113643201A (en) An adaptive non-local mean image denoising method
Mu et al. Low and non-uniform illumination color image enhancement using weighted guided image filtering
CN117252773A (en) Image enhancement method and system based on self-adaptive color correction and guided filtering
CN110211070A (en) A kind of low-luminance color image enchancing method based on local extremum
CN107203980B (en) Underwater target detection image enhancement method of self-adaptive multi-scale dark channel prior
CN110473152A (en) Based on the image enchancing method for improving Retinex algorithm
CN113160066A (en) Low-illumination image efficient enhancement method
CN111598886B (en) Pixel-level transmittance estimation method based on single image
CN112614063A (en) Image enhancement and noise self-adaptive removal method for low-illumination environment in building
CN114693548B (en) Dark channel defogging method based on bright area detection
Mi et al. A generalized enhancement framework for hazy images with complex illumination
CN116188339A (en) A Dark Vision Image Enhancement Method Based on Retinex and Image Fusion
CN110706180A (en) A method, system, device and medium for improving the visual quality of extremely dark images
CN114359083A (en) An adaptive preprocessing method for high dynamic thermal infrared images for interference environment
CN110796609B (en) Low-light image enhancement method based on scale-aware and detail enhancement model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20210723