CN106920237A - Based on empirical mode decomposition without with reference to full-colour image quality evaluating method - Google Patents
Based on empirical mode decomposition without with reference to full-colour image quality evaluating method Download PDFInfo
- Publication number
- CN106920237A CN106920237A CN201710130315.1A CN201710130315A CN106920237A CN 106920237 A CN106920237 A CN 106920237A CN 201710130315 A CN201710130315 A CN 201710130315A CN 106920237 A CN106920237 A CN 106920237A
- Authority
- CN
- China
- Prior art keywords
- image
- envelope surface
- minimum value
- image quality
- point set
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
This application discloses it is a kind of based on empirical mode decomposition without referring to full-colour image quality evaluating method, including the local maximum pixel point set drawn game portion minimum value pixel point set for calculating original pending image;Obtain maximum pixel and minimum value pixel envelope surface;Algebraic average is asked for as average curved surface to maximum pixel and minimum value pixel envelope surface;The intrinsic mode function component for causing that maximum pixel and minimum value pixel envelope surface reach unanimity is calculated, as ground floor detail view;New pending image is calculated, and calculates the n-th intrinsic mode function component and surplus result based on this, obtain n-layer detail view;The image information of n-layer detail view is counted, using based on the comprehensive estimation of score function image quality level for perceiving difference.Such scheme without artwork or artwork statistical parameter as reference, and taken into full account the evaluation indexes such as the brightness of each hierarchy chart picture, evaluation result and the subjective evaluation result of gained reach unanimity.
Description
Technical field
The application is related to image quality evaluation technical field, specifically, is related to a kind of nothing based on empirical mode decomposition
With reference to full-colour image quality evaluating method.
Background technology
Image quality evaluation is a research topic for classics, and its target is algorithm for design, provides the subjective feeling with people
Consistent evaluation of estimate.The mankind account for the mankind obtain gross information content 75% by the external information amount that vision is obtained, and image is
Human perception and the important information source of Machinery model identification, its quality play decision to the adequacy and accuracy of acquired information
The effect of property.The research of image quality evaluation has turned into one of basic technology of image information engineering.
The evaluation of current digital image quality can be divided into subjective assessment and objective evaluation.
Because the last observer of image is generally human eye, therefore, people initially study the subjectivity of picture quality
Evaluation test method.Subjective evaluation method be by contrived experiment, by constituted with statistic of classification meaning observer group to figure
As quality is evaluated.Comparing is typically considered accurately by subjective testing evaluation image quality and knot is reliably evaluated
Really.But subjective assessment test is not only cumbersome, time-consuming but also carries out fairly expensive, and also suffer from observer's specialty back of the body
The influence of the subjective factors such as scape, psychology and motivation, and not can be incorporated into and use in other algorithms.This is caused under many circumstances
All be inconvenient to carry out subjective testing.
Therefore, whether the objective evaluation of digital picture quality is increasingly paid attention to by people, from needing original reference
For the angle of image (being considered undistorted or with " perfection " quality), can be by image quality evaluating method point
It is three kinds:Full reference type (Full Reference, FR), partial reference type (Reduced Reference, RR), without reference type
(No Reference, NR).
FR picture quality methods need to provide complete original reference image information, are used as evaluating distorted image matter
The reference of amount.The characteristics of FR type image quality evaluating methods is clear mathematical sense, is easy to implement, and can capture pixel aspect
Slight distortion.RR image quality evaluating methods need not provide complete original reference image information, only using the original of part
Image information estimates the visually-perceptible quality of distorted image.NR algorithms can be divided into for the algorithm of type of distortion and based on machine
Two kinds of the algorithm of device study.The characteristics of this kind of method is that, without reference to image, flexibility is strong.In recent years, NR image quality evaluations
It is of interest by increasing scholar.
The research tendency of image quality evaluation algorithm is presented based on the simple objective evaluation algorithm conversion, objective be combined
Evaluation algorithms, and without with reference to algorithm start turn into study hotspot.And in recent years, the image quality evaluation of oriented mission demand
Algorithm gradually increases, and main cause is that the emphasis evaluated picture quality under different application scenarios is different.
Some occasions more focus on image portraying in detail, and some occasions then focus on image definition in structure.Furthermore, no
With evaluation method type of distortion different for compression, fuzzy, additive noise etc. and picture material (natural scene, personage, refer to
Line, woven bag etc.) susceptibility it is also different.
The content of the invention
In view of this, technical problems to be solved in this application there is provided it is a kind of based on empirical mode decomposition without reference
Full-colour image quality evaluating method, there is provided for the picture quality evaluation for obscuring this type of distortion.Without artwork or original
In the case that the statistical parameter of figure is as reference, using based on the comprehensive objective evaluation function for perceiving difference, after being decomposed through EMD
Each tomographic image detail view carries out statistical appraisal, with the actual mass level of correct perceptual image.
In order to solve the above-mentioned technical problem, the application has following technical scheme:
It is a kind of based on empirical mode decomposition without refer to full-colour image quality evaluating method, it is characterised in that including:
For original pending image, the local maximum pixel point set drawn game portion of the original pending image is calculated
Minimum value pixel point set;
Interpolation processing is carried out to the local maximum pixel point set and the local minimum pixel point set respectively,
Obtain maximum pixel envelope surface and minimum value pixel envelope surface;
Algebraic average is asked for for the maximum pixel envelope surface and the minimum value pixel envelope surface, is made
It is average curved surface;
According to average curved surface and original pending view data, calculate so that the maximum pixel envelope surface and
The intrinsic mode function component that the minimum value pixel envelope surface reaches unanimity, as the first intrinsic mode function point
Amount, as the 1st layer of detail view;
According to the first intrinsic mode function component and the original pending image, new pending image is calculated, with
The n-th intrinsic mode function component and surplus result are calculated based on new pending image, n-layer detail view, wherein n is got
≥2;
The image information of the n-layer detail view is counted, using based on the comprehensive estimation of score function picture quality for perceiving difference
Level.
Preferably, wherein:
For original pending image, the local maximum pixel point set drawn game portion of the original pending image is calculated
Minimum value pixel point set, further for:
For original pending image, maximum pixel and minimum value pixel, image are selected in 3x3 neighborhoods successively
Boundary extension predetermined width, obtains the local maximum pixel point set drawn game portion minimum value picture of the original pending image
Vegetarian refreshments set.
Preferably, wherein:
Interpolation processing is carried out to the local maximum pixel point set and the local minimum pixel point set respectively,
Obtain maximum pixel envelope surface and minimum value pixel envelope surface, further for:
By cubic spline interpolation method to the local maximum pixel point set and the local minimum pixel
Set carries out interpolation processing, obtains maximum pixel envelope surface and minimum value pixel envelope surface.
Preferably, wherein:
According to average curved surface and original pending view data, calculate so that the maximum pixel envelope surface and
The intrinsic mode function component that the minimum value pixel envelope surface reaches unanimity, as the first intrinsic mode function point
Amount, as ground floor detail view, further for:
After subtracting algebraic average curved surface mI (x, y) using original pending view data I (x, y), its result h is judged1(x,
Y) whether the condition of intrinsic mode function component is met, if it is not satisfied, then must be by h1(x, y) is used as new processed image, profit
Use h2(x, y)=h1(x,y)-m1I (x, y) repeats to ask for the process of average curved surface mI (x, y), after repeating k times, works as mkI(x,y)
Go to zero, maximum pixel envelope surface and minimum value pixel envelope surface reach unanimity, hk+1(x, y)=hk(x,y)-
mkI (x, y), the 1st intrinsic mode function component c1=hk+1(x, y), as the 1st layer of detail view.
Preferably, wherein:
According to the first intrinsic mode function component and the original pending image, new pending image is calculated, with
The n-th intrinsic mode function component and surplus result are calculated based on new pending image, n-layer detail view is got, enters one
Walk and be:
By the 1st intrinsic mode function component c1Separated from original pending image, with original pending image with
1st intrinsic mode function component c1Difference as new processed image, based on new pending image, is obtained successively
2nd, 3 ... n-layer detail views, until difference surplus stops in monotonic trend or during in default value scope, obtain n-layer detail view and
Surplus result.
Preferably, wherein:
The image information of the n-layer detail view is counted, using based on the comprehensive estimation of score function picture quality for perceiving difference
Level, further for:
Using the comprehensive evaluation function D for perceiving differenceGPImage quality level is estimated, wherein the comprehensive evaluation function for perceiving difference
DGPFor:
Wherein, the entropy of H representative images, G represents average gradient, DHPRepresentative image monochrome information, DHPBy the accumulative straight of image
Square figure distribution statistics are characterized, DHPIt is embodied as:
Wherein, n is the detail view number of plies after decomposing, and l is the number of greyscale levels of image, and C (l) is input picture accumulative histogram
Distribution function, C0L () is standard picture accumulative histogram distribution function.
Preferably, wherein:
The acquisition methods of the input picture accumulative histogram distribution function are:
If mlIt is number of pixels (0≤l≤L-1,0≤m with gray level l in imagel≤ m), L is gray level sum, m
It is total number of image pixels, then its Normalized Grey Level histogram probability density function is:
Thus input picture accumulative histogram distribution function is obtained:
Preferably, wherein:
The acquisition methods of the standard picture accumulative histogram distribution function are:
N-layer detail pictures accumulative histogram to being obtained after empirical mode decomposition method is decomposed is averaging, and is marked
Quasi- image histogram accumulative histogram distribution function:
Preferably, wherein:
The entropy H of described image is obtained by with minor function:
Wherein, P (l) is Normalized Grey Level histogram probability density function.
Preferably, wherein:
The average gradient G, obtains by with minor function:
Wherein,WithRespectively (x, y) puts gradient of the pixel grey scale on its row, column direction, M × N
It is image size.
Compared with prior art, method described herein, has reached following effect:
First, it is of the present invention based on empirical mode decomposition without refer to full-colour image quality evaluating method, in image matter
Amount evaluation field introduces empirical mode decomposition method, the characteristics of the method has adaptivity, can preferably retain original figure
The signal characteristic of picture, the accurate texture and marginal information for weighing image, has stronger parsing power to image quality level.
Second, it is of the present invention based on empirical mode decomposition without refer to full-colour image quality evaluating method, make use of base
In the objective evaluation function for comprehensively perceiving difference, entropy, average gradient, the brightness vision information of image are covered, fully have rated figure
The brightness of picture, gray scale and the inferior many index of levels of detail, the image quality evaluation result for obtaining have more persuasion.
3rd, it is of the present invention based on empirical mode decomposition without full-colour image quality evaluating method is referred to, without former
In the case that figure or artwork statistical parameter are as reference, the image information of each detail view after EMD is decomposed is counted, it evaluates knot
Really compared with the tradition evaluation method such as average, standard deviation, and accuracy more with consistency with the subjective vision sensing results of human eye.
Brief description of the drawings
Accompanying drawing described herein is used for providing further understanding of the present application, constitutes the part of the application, this Shen
Schematic description and description please does not constitute the improper restriction to the application for explaining the application.In the accompanying drawings:
Fig. 1 is the flow chart without reference full-colour image quality evaluating method based on empirical mode decomposition of the present invention;
Fig. 2 is the application implementation without reference full-colour image quality evaluating method based on empirical mode decomposition of the present invention
The flow chart of example.
Specific embodiment
Some vocabulary have such as been used to censure specific components in the middle of specification and claim.Those skilled in the art should
It is understood that hardware manufacturer may call same component with different nouns.This specification and claims are not with name
The difference of title is used as distinguishing the mode of component, but the difference with component functionally is used as the criterion distinguished.Such as logical
The "comprising" of piece specification and claim mentioned in is an open language, therefore should be construed to " include but do not limit
In "." substantially " refer to that in receivable error range, those skilled in the art can solve described in the range of certain error
Technical problem, basically reaches the technique effect.Additionally, " coupling " one word herein comprising it is any directly and indirectly electric property coupling
Means.Therefore, if a first device is coupled to a second device described in text, representing the first device can direct electrical coupling
The second device is connected to, or the second device is electrically coupled to indirectly by other devices or coupling means.Specification
Subsequent descriptions be implement the application better embodiment, so it is described description be for the purpose of the rule for illustrating the application,
It is not limited to scope of the present application.The protection domain of the application ought be defined depending on the appended claims person of defining.
Embodiment 1
It is shown in Figure 1 for it is herein described it is a kind of based on empirical mode decomposition without with reference to full-colour image quality evaluation side
The flow chart of method, this is based on including without reference full-colour image quality evaluating method for empirical mode decomposition:
Step 101, for original pending image, calculate original pending image local maximum pixel point set and
Local minimum pixel point set;
Step 102, interpolation processing is carried out to local maximum pixel point set drawn game portion minimum value pixel point set respectively,
Obtain maximum pixel envelope surface and minimum value pixel envelope surface;
Step 103, algebraic average is asked for for maximum pixel envelope surface and minimum value pixel envelope surface, made
It is average curved surface;
Step 104, according to average curved surface and original pending view data, calculate so that maximum pixel envelope is bent
The intrinsic mode function component that face and minimum value pixel envelope surface reach unanimity, as the first intrinsic mode function point
Amount, as the 1st layer of detail view;
Step 105, according to the first intrinsic mode function component and original pending image, calculate new pending figure
Picture, calculates the n-th intrinsic mode function component and surplus result based on new pending image, gets n-layer detail view,
Wherein n >=2;
The image information of step 106, statistics n-layer detail view, using based on the comprehensive estimation of score function image for perceiving difference
Quality level.
It is of the present invention based on empirical mode decomposition without refer to full-colour image quality evaluating method, in image quality evaluation
Field introduces empirical mode decomposition method, the characteristics of the method has adaptivity, can preferably retain the letter of original image
Number feature, the accurate texture and marginal information for weighing image, has stronger parsing power to image quality level.
The comprehensive difference that perceives based on empirical mode decomposition involved by the application is without with reference to full-colour image quality evaluating method
Distorted image mainly for vague category identifier carries out performance rating, such as image quality evaluation after recovered treatment.Empirical Mode
Any sophisticated signal can be decomposed into several intrinsic mode functions limited and with certain physical significance by state decomposition method
(Intrinsic Mode Functions, IMF) component.The method does not preset base (function), but criterion has been pre-selected,
Decomposed based on the characteristic size that signal is included in itself, obtained limited rank intrinsic mode function, thus suffer from comparing per rank IMF
Clear and definite physical significance.The method does not have stationarity and linear requirement to data, the characteristics of with adaptivity, can be preferably
Retain the feature of original signal.
In above-mentioned steps 101, for original pending image, the local maximum picture of the original pending image is calculated
Vegetarian refreshments set and local minimum pixel point set, further for:For original pending image, selected in 3x3 neighborhoods successively
Maximum pixel and minimum value pixel are selected, extension predetermined width at image boundary obtains the part of original pending image
Maximum pixel point set drawn game portion minimum value pixel point set.
In above-mentioned steps 102, local maximum pixel point set drawn game portion minimum value pixel point set is inserted respectively
Value treatment, obtain maximum pixel envelope surface and minimum value pixel envelope surface, further for:Inserted by cubic spline
Value method carries out interpolation processing to local maximum pixel point set drawn game portion minimum value pixel point set, obtains maximum pixel
Point envelope surface and minimum value pixel envelope surface.
In above-mentioned steps 104, according to average curved surface and original pending view data, calculate so that maximum pixel
The intrinsic mode function component that envelope surface and minimum value pixel envelope surface reach unanimity, as the first intrinsic mode
Function component, as ground floor detail view, further for:
After subtracting algebraic average curved surface mI (x, y) using original pending view data I (x, y), its result h is judged1(x,
Y) whether the condition of intrinsic mode function component is met, if it is not satisfied, then must be by h1(x, y) is used as new processed image, profit
Use h2(x, y)=h1(x,y)-m1I (x, y) repeats to ask for the process of average curved surface mI (x, y), after repeating k times, works as mkI(x,y)
Go to zero, maximum pixel envelope surface and minimum value pixel envelope surface reach unanimity, hk+1(x, y)=hk(x,y)-
mkI (x, y), the 1st intrinsic mode function component c1=hk+1(x, y), as the 1st layer of detail view.
In above-mentioned steps 105, according to the first intrinsic mode function component and original pending image, calculate and new wait to locate
Reason image, calculates the n-th intrinsic mode function component and surplus result based on new pending image, gets n-layer thin
Section figure, further for:
By the 1st intrinsic mode function component c1Separated from original pending image, with original pending image with
1st intrinsic mode function component c1Difference as new processed image, based on new pending image, is obtained successively
2nd, 3 ... n-layer detail views, until difference surplus stops in monotonic trend or during in default value scope, obtain n-layer detail view and
Surplus result.
In above-mentioned steps 106, the image information of the n-layer detail view is counted, using based on the comprehensive evaluation letter for perceiving difference
Number estimate image quality levels, further for:
Using the comprehensive evaluation function D for perceiving differenceGPImage quality level is estimated, wherein the comprehensive evaluation function for perceiving difference
DGPReferring to formula (1):
Wherein, the entropy of H representative images, G represents average gradient, DHPRepresentative image monochrome information, DHPBy the accumulative straight of image
Square figure distribution statistics are characterized, DHPSmaller, the brightness vision perceived effect of image is better, DHPIt is bigger, the brightness vision of image
Perceived effect is poorer.DHPIt is embodied as:
Wherein, n is the detail view number of plies after decomposing, and l is the number of greyscale levels of image, and C (l) is input picture accumulative histogram
Distribution function, C0L () is standard picture accumulative histogram distribution function.Histogram reflects the distribution feelings of image overall gray level
Condition.The image of good visual effect, its histogram can equably display image bloom, shadow and middle tone information, and density range
It is moderate.
DGPSame DHPEqually, DGPSmaller, treatment image is smaller with the gap of ideal image, and brightness perceives also better, image matter
Amount evaluating ability is also higher.Due to DGPInvolved evaluating does not need original reference image, therefore is without the objective of reference
Evaluation method.
Further, the acquisition methods of the input picture accumulative histogram distribution function are:
If mlIt is number of pixels (0≤l≤L-1,0≤m with gray level l in imagel≤ m), L is gray level sum, m
It is total number of image pixels, then its Normalized Grey Level histogram probability density function is:
Thus input picture accumulative histogram distribution function is obtained:
Further, the acquisition methods of the standard picture accumulative histogram distribution function are:
To being averaging by the n-layer detail pictures accumulative histogram obtained after empirical mode decomposition method (EMD) decomposition, obtain
To standard picture histogram accumulative histogram distribution function:
Further, gradation of image distribution is more uniform, and gray level level is more, and entropy is bigger, and the entropy H of image is by following
Function is obtained:
Wherein, P (l) is Normalized Grey Level histogram probability density function.
Further, the average gradient G, obtains by with minor function:
Wherein,WithRespectively (x, y) puts gradient of the pixel grey scale on its row, column direction, M × N
It is image size.Average gradient can reflect image to minor detail contrast ability to express.Average gradient is bigger, represents image layer
Secondary more, image is more clear, and contrast is better, conversely, image is fuzzyyer.
Embodiment 2
A kind of Application Example of the invention presented below, referring to Fig. 2, a kind of synthesis based on empirical mode decomposition is perceived
Differ from without reference full-colour image quality evaluating method, including:
Step 201, the pending image of input.
Step 202, the local maximum point set drawn game portion's minimum point set for seeking image.
Detailed process is:Maximum pixel and minimum value pixel, side are selected in 3 × 3 neighborhoods to pending image
Boundary sentences extension one fixed width, obtains local maximum point set drawn game portion's minimum point set of image.
Step 203, interpolation processing is carried out to local maximum point set and local minimizer set respectively, try to achieve maximum
Value point envelope surface and minimum point envelope surface.
Specific method is:Local maximum point set and local minimizer set are entered by cubic spline interpolation method
Row interpolation, obtains maximum of points envelope surface and minimum point envelope surface.
The algebraic average of step 204, maximizing point envelope surface and minimum point envelope surface, as average curved surface.
Step 205, subtract average with input image data.
Step 206, judge whether to meet every layer of iteration stopping condition.It is then to go to step 207;It is no, then with step 205
In result, go to step 202 further calculate.
Whether specific iteration stopping condition is the condition for meeting intrinsic mode function component, i.e. whether average curved surface tends to
0, whether maximum of points envelope surface and minimum point envelope surface reach unanimity.
Step 207, i-th layer of detail view that result is obtained as decomposition.
Step 208, this layer of information is subtracted from pending image.
Step 209, judge whether to meet image EMD and decompose termination condition.It is then to go to step 210;It is no, then by step 208
Result goes to step 201 as pending image, and then calculates i+1 layer detail view.
Specific EMD decompose termination condition be input picture with the difference surplus of detail pictures substantially in monotonic trend or
Very little, can be considered measurement error.
Step 210, obtain n-layer detail view and surplus result.
Gradation of image information after step 211, statistics EMD decomposition in each levels of detail.
Step 212, using based on the comprehensive estimation of score function image quality level for perceiving difference, specific picture quality is commented
Shown in valency function such as following formula (i.e. previously described formula (1)).
Wherein, wherein, the entropy of H representative images, G represents average gradient, DHPRepresentative image monochrome information, DHPBy image
Accumulative histogram distribution statistics are characterized, DHPSmaller, the brightness vision perceived effect of image is better, DHPIt is bigger, image it is bright
Degree visual perception is poorer.Computational methods to each variable in above formula refer to embodiment 1.
By various embodiments above, the beneficial effect that the application is present is:
First, it is of the present invention based on empirical mode decomposition without refer to full-colour image quality evaluating method, in image matter
Amount evaluation field introduces empirical mode decomposition method, the characteristics of the method has adaptivity, can preferably retain original figure
The signal characteristic of picture, the accurate texture and marginal information for weighing image, has stronger parsing power to image quality level.
Second, it is of the present invention based on empirical mode decomposition without refer to full-colour image quality evaluating method, make use of base
In the objective evaluation function for comprehensively perceiving difference, entropy, average gradient, the brightness vision information of image are covered, fully have rated figure
The brightness of picture, gray scale and the inferior many index of levels of detail, the image quality evaluation result for obtaining have more persuasion.
3rd, it is of the present invention based on empirical mode decomposition without full-colour image quality evaluating method is referred to, without former
In the case that figure or artwork statistical parameter are as reference, the image information of each detail view after EMD is decomposed is counted, it evaluates knot
Really compared with the tradition evaluation method such as average, standard deviation, and accuracy more with consistency with the subjective vision sensing results of human eye.
It should be understood by those skilled in the art that, embodiments herein can be provided as method, device or computer program
Product.Therefore, the application can be using the reality in terms of complete hardware embodiment, complete software embodiment or combination software and hardware
Apply the form of example.And, the application can be used and wherein include the computer of computer usable program code at one or more
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) is produced
The form of product.
Described above has shown and described some preferred embodiments of the application, but as previously described, it should be understood that the application
Be not limited to form disclosed herein, be not to be taken as the exclusion to other embodiment, and can be used for various other combinations,
Modification and environment, and can be in invention contemplated scope described herein, by above-mentioned teaching or the technology or knowledge of association area
It is modified.And the change and change that those skilled in the art are carried out do not depart from spirit and scope, then all should be in this Shen
Please be in the protection domain of appended claims.
Claims (10)
1. it is a kind of based on empirical mode decomposition without refer to full-colour image quality evaluating method, it is characterised in that including:
For original pending image, the local maximum pixel point set drawn game portion for calculating the original pending image is minimum
Value pixel point set;
Interpolation processing is carried out to the local maximum pixel point set and the local minimum pixel point set respectively, is obtained
Maximum pixel envelope surface and minimum value pixel envelope surface;
Algebraic average is asked for for the maximum pixel envelope surface and the minimum value pixel envelope surface, as equal
Value curved surface;
According to average curved surface and original pending view data, calculate so that the maximum pixel envelope surface and described
The intrinsic mode function component that minimum value pixel envelope surface reaches unanimity, as the first intrinsic mode function component,
As the 1st layer of detail view;
According to the first intrinsic mode function component and the original pending image, new pending image is calculated, with new
The n-th intrinsic mode function component and surplus result are calculated based on pending image, n-layer detail view, wherein n >=2 is got;
The image information of the n-layer detail view is counted, using based on the comprehensive estimation of score function image quality level for perceiving difference.
2. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 1
In for original pending image, the local maximum pixel point set drawn game portion for calculating the original pending image is minimum
Value pixel point set, further for:
For original pending image, maximum pixel and minimum value pixel, image boundary are selected in 3x3 neighborhoods successively
Place's extension predetermined width, obtains the local maximum pixel point set drawn game portion minimum value pixel of the original pending image
Set.
3. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 1
In, interpolation processing is carried out to the local maximum pixel point set and the local minimum pixel point set respectively, obtain
Maximum pixel envelope surface and minimum value pixel envelope surface, further for:
By cubic spline interpolation method to the local maximum pixel point set and the local minimum pixel point set
Interpolation processing is carried out, maximum pixel envelope surface and minimum value pixel envelope surface is obtained.
4. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 1
According to average curved surface and original pending view data, calculating so that the maximum pixel envelope surface and described
The intrinsic mode function component that minimum value pixel envelope surface reaches unanimity, as the first intrinsic mode function component,
As ground floor detail view, further for:
After subtracting algebraic average curved surface mI (x, y) using original pending view data I (x, y), its result h is judged1(x, y) is
The no condition for meeting intrinsic mode function component, if it is not satisfied, then must be by h1(x, y) as new processed image, using h2
(x, y)=h1(x,y)-m1I (x, y) repeats to ask for the process of average curved surface mI (x, y), after repeating k times, works as mkI (x, y) tends to
Zero, maximum pixel envelope surface and minimum value pixel envelope surface reach unanimity, hk+1(x, y)=hk(x,y)-mkI(x,
Y), the 1st intrinsic mode function component c1=hk+1(x, y), as the 1st layer of detail view.
5. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 1
In, according to the first intrinsic mode function component and the original pending image, new pending image is calculated, treated with new
Treatment image based on calculate the n-th intrinsic mode function component and surplus result, get n-layer detail view, further for:
By the 1st intrinsic mode function component c1Separated from original pending image, with original pending image and the 1st
Levy mode function component c1Difference as new processed image, based on new pending image, successively obtain the 2nd, 3 ...
N-layer detail view, stops when difference surplus is in monotonic trend or in default value scope, obtains n-layer detail view and surplus
As a result.
6. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 1
In, the image information of the n-layer detail view is counted, poor estimation of score function image quality level is perceived using based on comprehensive,
It is further:
Using the comprehensive evaluation function D for perceiving differenceGPImage quality level is estimated, wherein the comprehensive evaluation function D for perceiving differenceGPFor:
Wherein, the entropy of H representative images, G represents average gradient, DHPRepresentative image monochrome information, DHPBy the accumulative histogram of image
Distribution statistics are characterized, DHPIt is embodied as:
Wherein, n is the detail view number of plies after decomposing, and l is the number of greyscale levels of image, and C (l) is distributed for input picture accumulative histogram
Function, C0L () is standard picture accumulative histogram distribution function.
7. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 6
In the acquisition methods of the input picture accumulative histogram distribution function are:
If mlIt is number of pixels (0≤l≤L-1,0≤m with gray level l in imagel≤ m), L is gray level sum, and m is figure
As sum of all pixels, then its Normalized Grey Level histogram probability density function is:
Thus input picture accumulative histogram distribution function is obtained:
8. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 7
In the acquisition methods of the standard picture accumulative histogram distribution function are:
N-layer detail pictures accumulative histogram to being obtained after empirical mode decomposition method is decomposed is averaging, and obtains standard drawing
As histogram accumulative histogram distribution function:
9. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 7
In the entropy H of described image is obtained by with minor function:
Wherein, P (l) is Normalized Grey Level histogram probability density function.
10. existed without full-colour image quality evaluating method, its feature is referred to based on empirical mode decomposition according to claim 6
In the average gradient G is obtained by with minor function:
Wherein, ▽xI (x, y) and ▽yI (x, y) is respectively gradient of (x, y) point pixel grey scale on its row, column direction, and M × N is
Image size.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710130315.1A CN106920237A (en) | 2017-03-07 | 2017-03-07 | Based on empirical mode decomposition without with reference to full-colour image quality evaluating method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710130315.1A CN106920237A (en) | 2017-03-07 | 2017-03-07 | Based on empirical mode decomposition without with reference to full-colour image quality evaluating method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106920237A true CN106920237A (en) | 2017-07-04 |
Family
ID=59460244
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710130315.1A Pending CN106920237A (en) | 2017-03-07 | 2017-03-07 | Based on empirical mode decomposition without with reference to full-colour image quality evaluating method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106920237A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108428232A (en) * | 2018-03-20 | 2018-08-21 | 合肥工业大学 | A kind of blind appraisal procedure of cartoon image quality |
CN109754390A (en) * | 2018-12-11 | 2019-05-14 | 西北大学 | A kind of non-reference picture quality appraisement method based on mixing visual signature |
CN113012079A (en) * | 2021-05-25 | 2021-06-22 | 南京索安电子有限公司 | Low-brightness vehicle bottom image enhancement method and device and storage medium |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273787A1 (en) * | 2005-09-09 | 2008-11-06 | Qinetiq Limited | Automated Selection of Image Regions |
CN104103065A (en) * | 2013-04-10 | 2014-10-15 | 无锡金帆钻凿设备股份有限公司 | No-reference fuzzy image quality evaluation method based on singular value decomposition |
-
2017
- 2017-03-07 CN CN201710130315.1A patent/CN106920237A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080273787A1 (en) * | 2005-09-09 | 2008-11-06 | Qinetiq Limited | Automated Selection of Image Regions |
CN104103065A (en) * | 2013-04-10 | 2014-10-15 | 无锡金帆钻凿设备股份有限公司 | No-reference fuzzy image quality evaluation method based on singular value decomposition |
Non-Patent Citations (2)
Title |
---|
贺金平 等: "基于EMD的无参考图像清晰度评价方法", 《航天返回与遥感》 * |
高昆 等: "结合人眼视觉综合感知差与r-g 色域峰值偏差的彩色图像客观评价方法", 《北京理工大学学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108428232A (en) * | 2018-03-20 | 2018-08-21 | 合肥工业大学 | A kind of blind appraisal procedure of cartoon image quality |
CN108428232B (en) * | 2018-03-20 | 2019-07-19 | 合肥工业大学 | A kind of blind appraisal procedure of cartoon image quality |
CN109754390A (en) * | 2018-12-11 | 2019-05-14 | 西北大学 | A kind of non-reference picture quality appraisement method based on mixing visual signature |
CN109754390B (en) * | 2018-12-11 | 2023-04-07 | 西北大学 | No-reference image quality evaluation method based on mixed visual features |
CN113012079A (en) * | 2021-05-25 | 2021-06-22 | 南京索安电子有限公司 | Low-brightness vehicle bottom image enhancement method and device and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110046673B (en) | No-reference tone mapping image quality evaluation method based on multi-feature fusion | |
Qureshi et al. | Towards the design of a consistent image contrast enhancement evaluation measure | |
Fu et al. | A weighted variational model for simultaneous reflectance and illumination estimation | |
Fu et al. | A probabilistic method for image enhancement with simultaneous illumination and reflectance estimation | |
Provenzi et al. | Random spray retinex: a new retinex implementation to investigate the local properties of the model | |
Narvekar et al. | A no-reference image blur metric based on the cumulative probability of blur detection (CPBD) | |
Aydin et al. | Dynamic range independent image quality assessment | |
CN104134204B (en) | Image definition evaluation method and image definition evaluation device based on sparse representation | |
US9292910B2 (en) | Image processing apparatus, image processing method, program, printing medium, and recording medium | |
Wilson et al. | A new metric for grey-scale image comparison | |
CN102595185A (en) | Stereo image quality objective evaluation method | |
CN106920237A (en) | Based on empirical mode decomposition without with reference to full-colour image quality evaluating method | |
DE102018008275A1 (en) | Refine local parameterizations for the transmission of two-dimensional images to three-dimensional models | |
Gupta et al. | A HVS based perceptual quality estimation measure for color images | |
CN105976351B (en) | Stereo image quality evaluation method based on central offset | |
EP2983130A1 (en) | Image enhancement method and device | |
Guo et al. | Analysis of texture characteristics associated with visual complexity perception | |
CN106157264A (en) | Large area image uneven illumination bearing calibration based on empirical mode decomposition | |
Finlayson et al. | Lookup-table-based gradient field reconstruction | |
CN103955921B (en) | Image noise estimation method based on human eye visual features and partitioning analysis method | |
CN104408736A (en) | Characteristic-similarity-based synthetic face image quality evaluation method | |
Song et al. | Optimized tone mapping function for contrast enhancement considering human visual perception system | |
Li et al. | Stereoscopic 3D Image Retargeting Quality Assessment. | |
CN106548184A (en) | A kind of method and apparatus of adjustment illumination balance | |
CN109785418A (en) | The blinkpunkt of view-based access control model sensor model renders optimization algorithm |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170704 |
|
RJ01 | Rejection of invention patent application after publication |