CN109685742A - A kind of image enchancing method under half-light environment - Google Patents
A kind of image enchancing method under half-light environment Download PDFInfo
- Publication number
- CN109685742A CN109685742A CN201811639590.7A CN201811639590A CN109685742A CN 109685742 A CN109685742 A CN 109685742A CN 201811639590 A CN201811639590 A CN 201811639590A CN 109685742 A CN109685742 A CN 109685742A
- Authority
- CN
- China
- Prior art keywords
- image
- light
- brightness
- transformation
- color
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 230000009466 transformation Effects 0.000 claims abstract description 37
- 230000002708 enhancing effect Effects 0.000 claims abstract description 28
- 238000001914 filtration Methods 0.000 claims abstract description 8
- 210000001116 retinal neuron Anatomy 0.000 claims abstract description 8
- 230000002146 bilateral effect Effects 0.000 claims abstract description 7
- 238000006243 chemical reaction Methods 0.000 claims abstract description 6
- 238000000605 extraction Methods 0.000 claims abstract description 6
- 230000008569 process Effects 0.000 claims abstract description 6
- 230000009467 reduction Effects 0.000 claims abstract description 6
- 238000012935 Averaging Methods 0.000 claims abstract description 5
- 239000000284 extract Substances 0.000 claims abstract description 5
- 230000004438 eyesight Effects 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 6
- 230000010339 dilation Effects 0.000 claims description 3
- 238000011426 transformation method Methods 0.000 claims description 2
- 235000013399 edible fruits Nutrition 0.000 claims 1
- 230000001537 neural effect Effects 0.000 claims 1
- 210000001525 retina Anatomy 0.000 claims 1
- 230000035807 sensation Effects 0.000 claims 1
- 238000005286 illumination Methods 0.000 abstract description 9
- 238000012545 processing Methods 0.000 abstract description 5
- 238000004422 calculation algorithm Methods 0.000 description 13
- 230000000694 effects Effects 0.000 description 13
- 230000000007 visual effect Effects 0.000 description 8
- 230000003014 reinforcing effect Effects 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 238000011156 evaluation Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000011833 dog model Methods 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000000638 stimulation Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 241000282326 Felis catus Species 0.000 description 1
- 241001465754 Metazoa Species 0.000 description 1
- 101100285000 Neurospora crassa (strain ATCC 24698 / 74-OR23-1A / CBS 708.71 / DSM 1257 / FGSC 987) his-3 gene Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000003592 biomimetic effect Effects 0.000 description 1
- 210000004027 cell Anatomy 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 229910003460 diamond Inorganic materials 0.000 description 1
- 239000010432 diamond Substances 0.000 description 1
- 210000000609 ganglia Anatomy 0.000 description 1
- 210000002592 gangliocyte Anatomy 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006740 morphological transformation Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Classifications
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
- G06T2207/20028—Bilateral filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
Abstract
A kind of image enchancing method under half-light environment belongs to field of image processing;It is distorted in the presence of performance results after very serious enhancing, image outline is fuzzy, the low problem of color saturation;Including extracting the luminance component V in half-light image in HSV, the logarithmic transformation of global brightness is carried out, low-light (level) region in half-light image is carried out to the promotion of overall brightness;The neighborhood averaging brightness of pixel in current half-light image is obtained using the bilateral filtering under three Gauss model of receptive field of retinal neurons;It using top bottom cap conversion process, realizes and extracts function of the part compared with bright part from image background, top bottom cap transformation is recycled to carry out light and shade extraction and noise reduction to image;The present invention can not only show the details after half-light color image enhancement well, solve the problems, such as that illumination patterns are non-uniform, promote the overall contrast of image, be clearly outlined image detail, and color saturation is high.
Description
Technical field
The invention belongs to the image enchancing methods under field of image processing more particularly to a kind of half-light environment.
Background technique
It is well known that the image enhancement technique in image procossing occupies always very important work in picture research field
With the development of the technology of accompanying image enhancing in recent years, researcher proposes many effective image enchancing methods, such as small echo
Converter technique etc. achieves significant reinforcing effect, but these Enhancement Methods are because simple introducing mathematical method increases to image
In strong field, visual perception's characteristic of the mankind is not accounted for, performance results are distorted after there is very serious enhancing, image wheel
Exterior feature is fuzzy, the low problem of color saturation.
Summary of the invention
The present invention overcomes above-mentioned the deficiencies in the prior art, the image enchancing method under a kind of half-light environment is provided, it will be dark
Light color image rgb color space is converted to HSV space, carries out global brightness logarithmic transformation to half-light image;Secondly using view
Nethike embrane neuron receptive field three Gauss model carries out the setting contrast to image local edge;Finally according in morphology
Top cap transformation and bottom cap transformation, assisted the background extracting and noise reduction to illumination;The present invention can not only be shown well
Details after half-light color image enhancement solves the problems, such as that illumination patterns are non-uniform, promotes the overall contrast of image, makes figure
As details is clearly outlined, color saturation is high.
Technical solution of the present invention:
A kind of image enchancing method under half-light environment, comprising the following steps:
Step a, the luminance component V in half-light image in HSV is extracted, the logarithmic transformation of global brightness is carried out, meets the mankind
Low-light (level) region in half-light image is carried out the promotion of overall brightness by the subjective feeling of visual characteristic;
Step b, it based on using three Gauss model of the receptive field of retinal neurons, and carries out bilateral under the model
It filters to obtain the neighborhood averaging brightness of pixel in current half-light image;
Step c, according to the relationship between the brightness value of point and the average brightness of its neighborhood in present image, pass through human eye
Brightness feel to realize with the nonlinear characteristic of the logarithm of actual light intensity in step a to overall brightness enhancing image
Local contrast enhancing;
Step d, it using top bottom cap conversion process, realizes and extracts function of the part compared with bright part, then benefit from image background
Light and shade extraction and noise reduction are carried out to image with top bottom cap transformation, obtain final half-light enhancing image.
Further, the luminance component V extracted in half-light image in HSV is to convert HSV space for rgb color space,
Three components are obtained, are luminance component V, chrominance component H and saturation degree component S respectively;The chrominance component H and saturation degree point
Amount S is remained unchanged, and is handled luminance component V.
Further, the formula for converting HSV space for rgb color space is as follows:
Wherein, R, G, B respectively represent the pixel value of the rgb space of original half-light color image.
Further, three Gauss model of receptive field of the retinal neurons is as follows:
Wherein, A1, A2, A3Respectively indicate the peak factor in center, surrounding and edge, σ1, σ2, σ3Respectively indicate center, four
The scale parameter in week and edge.
Further, top bottom cap transform method, including the following steps:
Step d1, top cap transformation and the transformation of bottom cap are respectively defined as:
I "=(IB)-I (6)
Wherein, I is original image, and image after I ' converts for top cap, I " is the transformed image of bottom cap, and B is structural element,Indicate that original half-light image and structural element B carry out opening operation the final result, i.e., fine obtains Background
Picture;IB indicates original half-light image and structural element B carries out the result that closed operation obtains;Top cap is added on the original image
The result of transformation subtracts the transformation of bottom cap again can be improved the contrast of half-light image entirety;
Step d2,3 × 3 structural element of selector disc shape carries out morphology operations;Coarse scale structures element is by small size knot
Constitutive element is repeatedly expanded and is obtained, and structural element shape and dilation operation are as follows;
The present invention has the advantages that compared with the existing technology
The present invention provides the image enchancing methods under a kind of half-light environment, and half-light color image rgb color space is turned
It is changed to HSV space, global brightness logarithmic transformation is carried out to half-light image;Secondly three Gaussian mode of retinal neurons receptive field is used
Type carries out the setting contrast to image local edge;Finally according to the top cap transformation and the transformation of bottom cap in morphology, carry out
Assist the background extracting and noise reduction to illumination;The present invention can not only show the details after half-light color image enhancement well,
Solve the problems, such as that illumination patterns are non-uniform, moreover it is possible to which the overall contrast for promoting image is clearly outlined image detail, and color is full
It is high with degree.
Detailed description of the invention
Fig. 1 is flow chart of the present invention;
Fig. 2 is the three Gauss model comparison diagram of receptive field of retinal neurons;
Fig. 3 is effect contrast figure.
Specific embodiment
Below with reference to attached drawing, the present invention is described in detail.
Specific embodiment one
A kind of image enchancing method under half-light environment, as shown in Figure 1, comprising the following steps:
Step a, the luminance component V in half-light image in HSV is extracted, the logarithm for carrying out global brightness to luminance component becomes
It changes, for meeting the subjective feeling of human visual system, low-light (level) region in half-light image is carried out to the promotion of overall brightness;
Step b, it based on using three Gauss model of the receptive field of retinal neurons, and carries out bilateral under the model
It filters to obtain the neighborhood averaging brightness of pixel in current half-light image;
Step c, according to the relationship between the brightness value of point and the average brightness of its neighborhood in present image, pass through human eye
Brightness feel to realize with the nonlinear characteristic of the logarithm of actual light intensity in step a to overall brightness enhancing image
Local contrast enhancing;
Although step d, the first two steps make half-light image obtain good reinforcing effect, the background light and shade of image
Or it is bad, using top bottom cap conversion process, realizes and extract function of the part compared with bright part from image background, recycle top
Cap transformation in bottom carries out light and shade extraction and noise reduction to image, obtains final half-light enhancing image.
Specifically, since color image RGB color has very high color correlation, present implementation will not
Tri- color channels of RGB carry out the single treatment in each channel, are to convert HSV space for rgb color space, obtain three points
Amount, is luminance component V, chrominance component H and saturation degree component S respectively;The chrominance component H and saturation degree component S is kept not
Become, luminance component V is handled.
Specifically, the formula for converting HSV space for rgb color space is as follows:
Wherein, R, G, B respectively represent the pixel value of the rgb space of original half-light color image, and θ indicates the form and aspect of angle
Angle, π indicate 180 ° of angle.
Specifically, biological study shows that human eye vision has this non-linear particularity of logarithm to the reaction of brightness.People
Eye can be adjusted according to the variation of intensity of illumination in ambient enviroment, and human eye cannot work within the scope of one simultaneously, when
After human eye adapts under current average brightness, the receptive field of vision can be restricted.Such as in a dark room in
A lamp is lighted, it is also bigger to the stimulation of human eye than lighting ten lamps again in the present context.So in the picture, average brightness is big
Regional scope in gamma error, the visual experience of human eye is simultaneously insensitive.Human eye can fit in this current intensity of illumination
Should intensity brightness range.But under this intensity of illumination, human eye has restriction effect to the range, makes it
In certain brightness range hereinafter, leading to there is unidentifiable black to understand all stimulation brightness.
Specifically, the logarithmic transformation method of the global brightness includes that the darkness of half-light color image entirety is lower, image
In object it is most of covered by darkness, human eye subjectivity can not find out the profile of object.Logarithmic transformation is carried out to luminance component V,
Human eye vision be can satisfy to the subjective feeling of intensity of illumination, be completed at the same time the adjustment to image light and shade, the pressure of dynamic range
Contracting.Its formula is as follows.
In(x, y)=log (I (x, y)+1) (11)
In formula, I (x, y) is the value of luminance component V, and M is dynamic range of images maximum value 256, Ig(x, y) is to this
Luminance component is normalized.
In order to accurately describe the functional characteristic that the input-output of retinal ganglial cells receptive field shows, Rodieck
It is proposed that a kind of double gauss differential mode type, i.e. DOG model, its functional characteristic of the model energy quantitative description, with different sizes, same
The circle of heart overlapping shows the center of gangliocyte tradition receptive field and the distribution situation of perimeter region.However it is traditional
Receptive field is outer there is a wide range of area of disinthibiting, and Li et al. is analyzed original therein on the basis of DOG model deeply
Cause, and to propose three Gauss models on the basis of this, the Receptive Field of Visual Neurons institute of the model and same image stimulation animal
Obtained result is very identical.Image is handled using this model, the edge of image can not only be enhanced well, while it is also
The ingredient that frequency between half-light image low-to-medium altitude can efficiently be promoted finally meets the transmitting slowly varying information of brightness step
It needs.It on the basis of DOG model, is used by the 3rd Gaussian kernel of addition, area of disinthibiting on a large scale is represented with this.
Specifically, three Gauss model of receptive field of the retinal neurons is as follows:
Wherein, A1, A2, A3Respectively indicate the peak factor in center, surrounding and edge, σ1, σ2, σ3Respectively indicate center, four
The scale parameter in week and edge, (x, y) indicate the pixel of current location.
When the model is used to handle the brightness contrast marginal portion of half-light image, both fine it can must enhance marginal portion
Comparison, but can efficiently show to be crossed by heritage by the Yezhong heart periphery Antagonizing regional luminance comparison that filters out with
Half tone information.Three Gauss models for adopting receptive field in this embodiment, the characteristic in spatial domain are used
MATLAB simulation result as shown in Fig. 2, can clearly be embodied from the figure center stimulate excited protrusion, surrounding inhibit and
The impression characteristic distribution situation of the ganglia retinae of marginal portion excitement, the enhancing for half-light image have significant effect.
Specifically, bilateral filtering has the very strong performance kept to image border, while there are also the abilities of denoising.The party
Method is a kind of non-linear filtering method that space weight and similar weight all consider.Due to consider pixel numerical value and its
Reason of both on position: Euclidean distance and similarity degree between pixel.The bilateral filtering under three Gauss models is selected, it can
To fully demonstrate the information state at object edge position.The average brightness formula that image is obtained under this method is as follows:
In formula, GRWith GVFor space and the similar gaussian kernel function of two aspect of numerical value.
Specifically, after the brightness of Global treatment image, the contrast of image is still very poor, and image border profile is unclear.By
The relationship of the average brightness of current pixel brightness value and its contiguous range carries out the contrast enhancing of the part of image.Formula
It is as follows:
Ilin(x, y)=K (Ig(x,y)-I(x,y))+I(x,y) (17)
In formula, Ig(x, y) is image overall logarithmic transformation, and I (x, y) is the neighborhood averaging brightness of current pixel point, by
Three Gauss model bilateral filterings obtain.K is proportionality coefficient.
Specifically, after after half-light color image RGB conversion HSV color space to the processing of V luminance component, by global brightness
After enhancing and local luminance enhancing, the color of image is not restored.Shown in the following formula of method using color post-equalization.
In formula, Ij(x, y) respectively corresponds r, g, b three primary color components in original half-light image;SjAfter (x, y) is enhancing
Corresponding r, g, b three primary color components of color image.
Top cap transformation is a kind of important method of morphological transformation in image analysis processing, can be completed to wide area
It is interior to the single extraction compared with dark background, local brighter areas is extracted from background so as to can effectively further realize
Function.And the effect of bottom cap transformation is then on the contrary.Specifically, top bottom cap transform method, including the following steps:
Step d1, top cap transformation and the transformation of bottom cap are respectively defined as:
I "=(IB)-I (20)
Wherein, I is original image, and image after I ' converts for top cap, I " is the transformed image of bottom cap, and B is structural element,Indicate that original half-light image and structural element B carry out opening operation the final result, i.e., fine obtains Background
Picture;IB indicates original half-light image and structural element B carries out the result that closed operation obtains;The process of opening operation and closed operation
As shown in Figure 4;The selection of B generates critically important effect to image enhancement work.Top cap transformation has the work of similar high-pass filtering
With image detail part can be highlighted, and cap transformation in bottom then has a characteristic of low-pass filtering, prominent image is connected target
Boundary between position is pushed up this characteristic between the transformation of bottom cap and is used in combination in the research of half-light image enhancement, can be incited somebody to action
The target and background of half-light image obtain certain stretching, highlight the information of target Yu details position.On the original image plus
The result of upper top cap transformation subtracts the transformation of bottom cap again can efficiently improve the contrast of half-light image entirety very much;
The size and shape of selection for structural element B, operation and structural element in morphology have close pass
System, constructs different structural elements, can obtain different as a result, the effect of obtained half-light enhancing image is also different.Knot
Constitutive element B wants bounded geometrically must be simpler than original image.Structural element can be divided into different shapes, there is disk
Shape, diamond shape, hexagon etc..Since disc-shaped structure element has the characteristics that each to symmetrical, and 3 × 3 and 5 × 5 structural elements are the most
Common, wherein the arithmetic speed of 3 × 3 structural elements is fast, testing result is more fine, so 3 × 3 structure of selector disc shape
Element carries out morphology operations.After the shape of structural element has determined, the size of structural element is then to play enhancing result
An important factor for.The structural element of small size is on the weak side to the noise removal capability of image, but has to good edge details and detect energy
Power.Large-sized structural element noise removal capability is very strong, but image border detected seems relatively thick.
Step d2,3 × 3 structural element of selector disc shape carries out morphology operations;Coarse scale structures element is by small size knot
Constitutive element is repeatedly expanded and is obtained, and structural element shape and dilation operation are as follows;
Wherein, b indicates structural element, bnIndicate the coarse scale structures element expanded by n times.
Specific embodiment two
Some research workers carefully analyze research by the biological characteristic to human visual system, are applied to figure
In image intensifying field, the Enhancement Method based on Retinex theory is proposed, this method can enhance image in high dynamic range,
It attracts attention in recent years.However, major part Retinex method can obtain preferable image enhancement visual effect, but still it is universal
It needs to be respectively processed tri- components of RGB in image, had both been easy to cause the cross-color of image in this way, and due to
It is completed at the same time local contrast enhancing and dynamic range compression, so that the enhancing process of whole image is difficult to control.By deeply
Have studied biomimetic colour image enchancing method, it is found that the image enchancing method has preferable vision enhancement effect, but due to it
Ineffective to some BORDER PROCESSINGs in the picture, so that overall contrast is not high, picture seems soft edge.Pass through research
Top cap transformation and the transformation of bottom cap, it was found that it is fairly obvious for the light and shade abstraction function effect of details and background, then by two
Kind method combines, and enhances the SSR (Single-Scale in algorithm with the histogram equalization of current hot topic and Retinex
Retinx), MSRCR (Multi-Scale Retinex with Color Restoration) is compared.
On the basis of specific embodiment one, present embodiment has chosen four groups of comparison photos as test sample, choosing
The photo taken is the picture shot under subdued light conditions.A group image is potted plant, and b group is kitten, and c group is lake water, and d group is trip
Happy field.The method to compare with the present invention is histogram equalization, the SSR in Retinex algorithm, in Retinex algorithm
Tri- kinds of methods of MSRCR.Picture size employed in four groups of images is respectively 700 × 500,950 × 700,400 × 700,600
× 900 colored jpg formats.Test comparison chart is as shown in Figure 3;It (a1) is original image and its grey level histogram;(b1) be original image and
Its grey level histogram;It (c1) is original image and its grey level histogram;(d1) be original image and its grey level histogram;It (a2) is histogram
Scheme its balanced grey level histogram;It (a3) is SSR and its grey level histogram;It (a4) is MSRCR and its grey level histogram;(a5) it is
Invention and its grey level histogram herein;It (b2) is its grey level histogram of histogram equalization;It (b3) is SSR and its grey level histogram;
It (b4) is MSRCR and its grey level histogram;It (b5) is the present invention and its grey level histogram;It (c2) is histogram equalization and its ash
Spend histogram;It (c3) is SSR and its grey level histogram;It (c4) is MSRCR and its grey level histogram;(c5) be herein invention and
Its grey level histogram;It (d2) is histogram equalization and its grey level histogram;It (d3) is SSR and its grey level histogram;(d4) it is
MSRCR and its grey level histogram;(d5) be herein invention and its grey level histogram.
The comparison of original image and various Enhancement Method reinforcing effects, the direct feel of human eye vision are as follows: (a1), (b1),
(c1), (d1) original image is whole partially dark, and contour of object is unclear in figure, and local detail can not be observed directly by human eye, image
Saturation degree is poor.(a2), (b2), (c2), (d2) histogram equalization improve the overall contrast of image, but the details of image
Part is not exhibited by, and image is there are larger noise, cross-color, and image saturation is poor.(a3),(b3),(c3), (d3)
SSR algorithm makes the phenomenon that image is whole partially white, and there are excessive enhancings, and subjective vision effect is poor, level of detail fuzzy, image
Color distortion.(a4), (b4), (c4), (d4) MSRCR algorithm have preferable holding, image vision to the color of enhancing image
Effect is preferable, but picture saturation degree is low.(a5), the half-light image that (b5), (c5), (d5) present embodiment enhance, image pair
Higher than degree, leaf, cat, the clarity of detail of the scenery such as chair, sharp outline is rich in color, and image is lively, and clarity is high.It is logical
Effect picture comparison is crossed, the outdoor scene that can be derived that the half-light enhancing image that present embodiment obtains can be experienced with human eye more connects
Closely, details in kind is more prominent, and picture becomes apparent from, and image is vividly rich in color, the better quality of picture.
Can clearly it be found out by (a2), (b2), (c2), (d2) grey level histogram, histogram equalization algorithm is to image ash
Degree stretching is obvious, and intensity profile covers 0-255 range, and the probability density distribution of image gray levels can be made to be similar to
Even distribution, to improve image overall contrast ratio and enhance details.By (a3), (b3), (c3), (d3), (a4), (b4),
(c4), (d4) can be seen that SSR, MSRCR algorithm enhance the lower pixel of gray value, and gray value becomes larger after enhancing, so that
The pixel quantity of image low ash degree tails off.By (a5), (b5), (c5), (d5) histogram it can be seen to gray scale stretching ratio
More apparent, three peaks in former histogram retain obviously, retain the information of some gray scales preferable.
In order to verify effectiveness of the invention, luminance mean value, contrast is selected to promote index, Y-PSNR thus
PSNR3 index as present embodiment evaluation criterion.Luminance mean value index definition of the invention is to become RGB image
HIS color space is changed to, take to luminance component I therein the operation of mean value.
The definition of contrast promotion index:
3 × 3 fritters are divided the image into, C is expressed as the mean value of 3 × 3 whole small images contrasts, and fritter contrast is fixed
Justice are as follows:
In formula, max is the maximum value of small images gray value;Min is the minimum value of small images gray value.In subscript
Original and proposed symbol respectively indicates half-light original image and its enhanced image.
Evaluation criterion based on pixel difference, Y-PSNR PSNR.MSE indicates half-light image X and the input of enhancing
The mean square error of original half-light image Y, the calculating of MSE are shown in formula (10) that the H and W in formula are expressed as in image
Height and width, (i, j) indicate the pixel value at image midpoint.PSNR is from MSE calculating, it is used to a measurement image fault
Or noise level, PSNR is bigger, indicates that distortion and noise are smaller.Formula (11) are shown in the calculating of PSNR.PSNR index is most general
All over a kind of objective method for measurement of the criticism image quality used.
1 data of table compare
2 data of table compare
3 data of table compare
4 data of table compare
The objective indicator evaluation criterion of the image property for original image and after 4 kinds of method enhancings are handled of table 1- table 4.From bright
It spends from the point of view of mean angular, 4 kinds of methods all have significant reinforcing effect to original half-light image, improve the brightness of image.This
Invention luminance mean value is numerically lower than other enhancing algorithms.The experimental data in terms of contrast promotes index, histogram are equal
Weighing apparatusization and contrast of the invention promote that index is maximum, but the image effect that intuitively reflects of histogram equalization and bad, uncomfortable
The reinforcing effect of half-light is closed, because there is no go to consider the algorithm from the mechanism of image objects and the visual characteristic of human eye.And this
The contrast of invention promotes index compared with SSR and MSRCR algorithm, and numerical value is higher, there is good contrast reinforcing effect.
Available from Y-PSNR PSNR index, numerical value of the invention is larger, and the distortion of image compares it with noise
It is smaller for his 3 kinds of algorithms.To sum up to half-light enhancing subjective assessment and to the test data of MATLAB platform it is objective comparison come
It says, the present invention is better than the image enchancing method of other a few class hot topics.
The present invention is low for contrast in half-light low-light (level) image, and the fuzzy problem of image outline details proposes top bottom
The bionical algorithm for image enhancement of cap transformation.The present invention refers to three Gauss model of neuron receptive field, makes which increase image sides
The contrast of edge.Top bottom cap transformation, extraction and image denoising to image background are introduced simultaneously.The present invention and histogram
Scheme equilibrium, SSR, MSRCR algorithm carry out Experimental comparison, from objectively evaluating index luminance mean value, contrast promotion index and subjectivity
Visual experience carry out evaluation result.It confirms, half-light image enchancing method of the invention has preferably the contrast promotion of image
Effect, image grayscale range are stretched, and image outline detail section has more obvious performance, have in practical applications very strong
Researching value.
Claims (5)
1. the image enchancing method under a kind of half-light environment, which comprises the following steps:
Step a, the luminance component V in half-light image in HSV is extracted, the logarithmic transformation of global brightness is carried out, meets human vision
Low-light (level) region in half-light image is carried out the promotion of overall brightness by the subjective feeling of characteristic;
Step b, based on using three Gauss model of the receptive field of retinal neurons, and the bilateral filtering under the model is carried out
To obtain the neighborhood averaging brightness of pixel in current half-light image;
Step c, according to the relationship between the brightness value of point and the average brightness of its neighborhood in present image, pass through the master of human eye
It is right to the part of overall brightness enhancing image in step a to realize to see the nonlinear characteristic of the logarithm of brightness sensation and actual light intensity
Enhance than degree;
Step d, it using top bottom cap conversion process, realizes and extracts function of the part compared with bright part from image background, recycle top
Cap transformation in bottom carries out light and shade extraction and noise reduction to image, obtains final half-light enhancing image.
2. the image enchancing method under a kind of half-light environment according to claim 1, which is characterized in that extract in half-light image
Luminance component V in HSV is to convert HSV space for rgb color space, obtains three components, is luminance component V, color respectively
Adjust component H and saturation degree component S;The chrominance component H and saturation degree component S are remained unchanged, and are handled luminance component V.
3. the image enchancing method under a kind of half-light environment according to claim 2, which is characterized in that described by rgb color
Spatial transformation is that the formula of HSV space is as follows:
Wherein, R, G, B respectively represent the pixel value of the rgb space of original half-light color image.
4. the image enchancing method under a kind of half-light environment according to claim 1, which is characterized in that the retina neural
Three Gauss model of receptive field of member is as follows:
Wherein, A1, A2, A3Respectively indicate the peak factor in center, surrounding and edge, σ1, σ2, σ3Respectively indicate center, surrounding and
The scale parameter at edge.
5. the image enchancing method under a kind of half-light environment according to claim 1, which is characterized in that the top bottom cap transformation
Method, including the following steps:
Step d1, top cap transformation and the transformation of bottom cap are respectively defined as:
I "=(IB)-I (6)
Wherein, I is original image, and image after I ' converts for top cap, I " is the transformed image of bottom cap, and B is structural element,
Indicate that original half-light image and structural element B carry out opening operation the final result, i.e., fine obtains background image;I·B
Indicate that original half-light image and structural element B carry out the result that closed operation obtains;On the original image plus the knot of top cap transformation
Fruit subtracts the transformation of bottom cap again can be improved the contrast of half-light image entirety;
Step d2,3 × 3 structural element of selector disc shape carries out morphology operations;Coarse scale structures element is by small scale structures member
Element is repeatedly expanded and is obtained, and structural element shape and dilation operation are as follows;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811639590.7A CN109685742A (en) | 2018-12-29 | 2018-12-29 | A kind of image enchancing method under half-light environment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811639590.7A CN109685742A (en) | 2018-12-29 | 2018-12-29 | A kind of image enchancing method under half-light environment |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109685742A true CN109685742A (en) | 2019-04-26 |
Family
ID=66191308
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811639590.7A Pending CN109685742A (en) | 2018-12-29 | 2018-12-29 | A kind of image enchancing method under half-light environment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109685742A (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110415188A (en) * | 2019-07-10 | 2019-11-05 | 首都师范大学 | A kind of HDR image tone mapping method based on Multiscale Morphological |
CN110458761A (en) * | 2019-06-25 | 2019-11-15 | 千奥星科南京生物科技有限公司 | A kind of brain medicine fluorescence video image restorative procedure |
CN110738619A (en) * | 2019-10-15 | 2020-01-31 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
CN111179216A (en) * | 2019-12-03 | 2020-05-19 | 中国地质大学(武汉) | Crop disease identification method based on image processing and convolutional neural network |
CN111445419A (en) * | 2020-04-09 | 2020-07-24 | 南京图格医疗科技有限公司 | Medical endoscope image enhancement method based on mathematical morphology |
CN111968063A (en) * | 2020-09-07 | 2020-11-20 | 北京凌云光技术集团有限责任公司 | Morphological image filtering device and method |
GB2588674A (en) * | 2019-11-01 | 2021-05-05 | Apical Ltd | Image processing |
WO2021114564A1 (en) * | 2019-12-10 | 2021-06-17 | 南京莱斯电子设备有限公司 | Enhancement method for low-contrast infrared image |
CN113658280A (en) * | 2021-08-20 | 2021-11-16 | 平安科技(深圳)有限公司 | Data augmentation method, device, equipment and storage medium based on artificial intelligence |
CN113822809A (en) * | 2021-03-10 | 2021-12-21 | 无锡安科迪智能技术有限公司 | Dim light enhancement method and system |
CN114697584A (en) * | 2020-12-31 | 2022-07-01 | 杭州海康威视数字技术股份有限公司 | Image processing system and image processing method |
CN115100077A (en) * | 2022-07-25 | 2022-09-23 | 深圳市安科讯实业有限公司 | Novel image enhancement method and device |
CN115760629A (en) * | 2022-11-25 | 2023-03-07 | 黑龙江工程学院 | Low-illumination image enhancement method |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853490A (en) * | 2010-04-21 | 2010-10-06 | 中国科学院半导体研究所 | Bionic image restoration method based on human visual characteristics |
CN103971330A (en) * | 2013-02-05 | 2014-08-06 | 腾讯科技(深圳)有限公司 | Image enhancing method and device |
CN106960424A (en) * | 2017-03-31 | 2017-07-18 | 上海澜澈生物科技有限公司 | Tubercle bacillus image segmentation and identification method and device based on optimized watershed algorithm |
-
2018
- 2018-12-29 CN CN201811639590.7A patent/CN109685742A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853490A (en) * | 2010-04-21 | 2010-10-06 | 中国科学院半导体研究所 | Bionic image restoration method based on human visual characteristics |
CN103971330A (en) * | 2013-02-05 | 2014-08-06 | 腾讯科技(深圳)有限公司 | Image enhancing method and device |
CN106960424A (en) * | 2017-03-31 | 2017-07-18 | 上海澜澈生物科技有限公司 | Tubercle bacillus image segmentation and identification method and device based on optimized watershed algorithm |
Non-Patent Citations (3)
Title |
---|
G SARAVANAN_ ET AL: "Real Time Implementation of RGB to HSV/HSI/HSL and Its Reverse Color Space Models", 《INTERNATIONAL CONFERENCE ON COMMUNICATION AND SIGNAL PROCESSING》 * |
段彦华 等: "眼底图像中硬性渗出物检测算法", 《北京生物医学工程》 * |
金小贤 等: "一种基于视觉特性的仿生图像增强算法", 《计算机辅助设计与图形学学报》 * |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110458761A (en) * | 2019-06-25 | 2019-11-15 | 千奥星科南京生物科技有限公司 | A kind of brain medicine fluorescence video image restorative procedure |
CN110415188A (en) * | 2019-07-10 | 2019-11-05 | 首都师范大学 | A kind of HDR image tone mapping method based on Multiscale Morphological |
CN110738619B (en) * | 2019-10-15 | 2022-03-01 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
CN110738619A (en) * | 2019-10-15 | 2020-01-31 | 西南大学 | Image enhancement method based on bionic self-adaptive memristor cell neural network |
GB2588674A (en) * | 2019-11-01 | 2021-05-05 | Apical Ltd | Image processing |
US11823361B2 (en) | 2019-11-01 | 2023-11-21 | Arm Limited | Image processing |
GB2588674B (en) * | 2019-11-01 | 2022-02-02 | Apical Ltd | Image processing |
CN111179216A (en) * | 2019-12-03 | 2020-05-19 | 中国地质大学(武汉) | Crop disease identification method based on image processing and convolutional neural network |
CN111179216B (en) * | 2019-12-03 | 2023-03-28 | 中国地质大学(武汉) | Crop disease identification method based on image processing and convolutional neural network |
WO2021114564A1 (en) * | 2019-12-10 | 2021-06-17 | 南京莱斯电子设备有限公司 | Enhancement method for low-contrast infrared image |
CN111445419A (en) * | 2020-04-09 | 2020-07-24 | 南京图格医疗科技有限公司 | Medical endoscope image enhancement method based on mathematical morphology |
CN111968063A (en) * | 2020-09-07 | 2020-11-20 | 北京凌云光技术集团有限责任公司 | Morphological image filtering device and method |
CN111968063B (en) * | 2020-09-07 | 2024-01-26 | 凌云光技术股份有限公司 | Morphological image filtering device and method |
CN114697584A (en) * | 2020-12-31 | 2022-07-01 | 杭州海康威视数字技术股份有限公司 | Image processing system and image processing method |
CN114697584B (en) * | 2020-12-31 | 2023-12-26 | 杭州海康威视数字技术股份有限公司 | Image processing system and image processing method |
CN113822809A (en) * | 2021-03-10 | 2021-12-21 | 无锡安科迪智能技术有限公司 | Dim light enhancement method and system |
CN113822809B (en) * | 2021-03-10 | 2023-06-06 | 无锡安科迪智能技术有限公司 | Dim light enhancement method and system thereof |
CN113658280B (en) * | 2021-08-20 | 2023-07-04 | 平安科技(深圳)有限公司 | Data augmentation method, device, equipment and storage medium based on artificial intelligence |
CN113658280A (en) * | 2021-08-20 | 2021-11-16 | 平安科技(深圳)有限公司 | Data augmentation method, device, equipment and storage medium based on artificial intelligence |
CN115100077A (en) * | 2022-07-25 | 2022-09-23 | 深圳市安科讯实业有限公司 | Novel image enhancement method and device |
CN115100077B (en) * | 2022-07-25 | 2022-11-04 | 深圳市安科讯实业有限公司 | Image enhancement method and device |
CN115760629A (en) * | 2022-11-25 | 2023-03-07 | 黑龙江工程学院 | Low-illumination image enhancement method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109685742A (en) | A kind of image enchancing method under half-light environment | |
CN106683080B (en) | A kind of retinal fundus images preprocess method | |
CN105046663B (en) | A kind of adaptive enhancement method of low-illumination image for simulating human visual perception | |
CN103295191B (en) | Multiple scale vision method for adaptive image enhancement and evaluation method | |
CN107798661B (en) | Self-adaptive image enhancement method | |
CN109191390A (en) | A kind of algorithm for image enhancement based on the more algorithm fusions in different colours space | |
Yang et al. | An adaptive method for image dynamic range adjustment | |
CN110047051A (en) | A kind of non-uniform lighting colour-image reinforcing method | |
CN109872285A (en) | A kind of Retinex low-luminance color image enchancing method based on variational methods | |
CN106558031B (en) | A kind of image enchancing method of the colored eyeground figure based on imaging model | |
CN108257099A (en) | A kind of adaptive infrared image enhancing method of view-based access control model contrast resolution | |
CN108288258A (en) | A kind of low-quality images Enhancement Method under severe weather conditions | |
CN111968041A (en) | Self-adaptive image enhancement method | |
CN111476725A (en) | Image defogging enhancement algorithm based on gradient domain oriented filtering and multi-scale Retinex theory | |
CN101853490A (en) | Bionic image restoration method based on human visual characteristics | |
CN110675351B (en) | Marine image processing method based on global brightness adaptive equalization | |
CN113222877B (en) | Infrared and visible light image fusion method and application thereof in airborne photoelectric video | |
CN103839245B (en) | The Retinex colour-image reinforcing method at night of Corpus--based Method rule | |
CN105426847A (en) | Nonlinear enhancing method for low-quality natural light iris images | |
Zhuang et al. | Image enhancement by deep learning network based on derived image and Retinex | |
CN110246086A (en) | The image processing method and its image processing system of color image enhancement | |
Meylan | Tone mapping for high dynamic range images | |
CN104915934A (en) | Grayscale image enhancement method based on retina mechanism | |
Ma et al. | A new algorithm for backlight image enhancement | |
CN108416788B (en) | Edge detection method based on receptive field and light and shade adaptability thereof |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190426 |