CN110378893A - Image quality evaluating method, device and electronic equipment - Google Patents
Image quality evaluating method, device and electronic equipment Download PDFInfo
- Publication number
- CN110378893A CN110378893A CN201910674392.2A CN201910674392A CN110378893A CN 110378893 A CN110378893 A CN 110378893A CN 201910674392 A CN201910674392 A CN 201910674392A CN 110378893 A CN110378893 A CN 110378893A
- Authority
- CN
- China
- Prior art keywords
- target area
- image
- value
- distortion factor
- target
- 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.)
- Granted
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)
- Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
Abstract
The present invention provides a kind of image quality evaluating method, device and electronic equipments, wherein this method comprises: target image to be divided into the multiple regions block of default size;The sharpness value for calculating multiple regions block, obtains calculated result;According to calculated result, target area is extracted from multiple regions block;Extract the characteristic value of target area;According to the characteristic value of target area, the quality evaluation parameter of target image is determined.After target image is divided into multiple regions block by this method, target area is extracted according to the sharpness value of region unit, characteristic value is extracted from target area and determines the quality evaluation parameter of target image.The application difficulty that image quality evaluation can be mitigated, increases the accuracy of image quality evaluation.
Description
Technical field
The present invention relates to technical field of image processing, more particularly, to a kind of image quality evaluating method, device and electronics
Equipment.
Background technique
In the related technology, image quality evaluation is divided into subjective assessment and objectively evaluates two kinds.Wherein, subjective assessment refers to
Assessment picture quality is directly observed with human eye, it is not only time-consuming but also with many uncontrollable factors.It objectively evaluates based on calculating
Model evaluates image, objectively evaluates and is broadly divided into three classes: full reference image quality appraisement, half reference picture quality are commented
Valence, non-reference picture quality appraisement.Full reference picture quality requires harshness in practical application.Half reference image quality appraisement drop
Low application requirement, but also need original image portion feature.Non-reference picture quality appraisement does not need original image auxiliary, main
It is divided into tailored version image quality evaluation algorithm and universal image quality evaluation algorithm.In practical applications, tailored version is evaluated
Algorithm can not adapt to different type of distortion, using being very limited;And the accuracy of universal quality evaluation algorithm need
It improves.
Summary of the invention
In view of this, the purpose of the present invention is to provide a kind of image quality evaluating method, device and electronic equipment, to subtract
The application difficulty of small picture quality method increases the accuracy of picture quality method.
In a first aspect, the embodiment of the invention provides a kind of image quality evaluating methods, comprising: target image to be divided into
The multiple regions block of default size;The sharpness value for calculating multiple regions block, obtains calculated result;According to calculated result, from multiple
Target area is extracted in region unit;Extract the characteristic value of target area;According to the characteristic value of target area, target image is determined
Quality evaluation parameter.
In preferred embodiments of the present invention, the sharpness value of above-mentioned calculating multiple regions block, the step of obtaining calculated result,
Include: that the sharpness value of each pixel in current region block is calculated according to following formula for each region unit:By pixel each in current region block
The sum of sharpness value is determined as current region block sharpness value;Wherein, I (i, j) is the pixel coordinate in current region block;σ (i,
J) sharpness value for being pixel I (i, j);ωk.lIt is the discrete two-dimensional Gaussian function coefficient using k and l as parameter;K is value model
Enclose any integer value between-K to K;Any integer value of the l between value range-L to L;K and L is preset fixation
Value.
It is above-mentioned according to calculated result in preferred embodiments of the present invention, target area is extracted from multiple regions block
Step, comprising: according to the sequence that sharpness value is descending, multiple regions block is ranked up, ranking results are obtained;It is tied from sequence
First region BOB(beginning of block) in fruit, extracts the region unit of continuous specified quantity;The region unit extracted is determined as target
Region.
In preferred embodiments of the present invention, the step of the characteristic value of said extracted target area, comprising: to target area
Carry out Fuzzy Processing, the target area after obtaining Fuzzy Processing;The characteristic value of target area is calculated by following formula:Wherein, Δ t is the characteristic value of target area;N is the number of target area;I be value range 0 to N it
Between any integer value;F is that preset change degree calculates function;PiFor target area, P 'iFor the target area after Fuzzy Processing.
In preferred embodiments of the present invention, the above-mentioned characteristic value according to target area determines that the quality of target image is commented
The step of valence parameter, comprising: the characteristic value of target area is inputted in preset weighted model, the matter of target image is exported
Measure evaluation parameter;Weighted model is established according to the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor.
In preferred embodiments of the present invention, above-mentioned weighted model is established by following steps: being based on preset training set
Determine the standard evaluation score of training image and training image;Calculate the fuzzy distortion factor of training image, noise distortion degree and
The blocking artifact distortion factor;The fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor are normalized;According to training image
Standard evaluation score, adjust and obscure the distortion factor, noise distortion degree and the corresponding weight of the blocking artifact distortion factor in weighted model;After
It is continuous to execute the step of training image is determined based on preset training set, until the evaluation score and standard that are obtained according to weight calculation
The error of evaluation score within a preset range, obtains final weighted model.
In preferred embodiments of the present invention, the above-mentioned standard evaluation score according to training image is adjusted in initial model
The step of fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor corresponding weight, comprising: commented according to the standard of training image
Valence score calculates the fuzzy distortion factor, the weight of noise distortion degree and the blocking artifact distortion factor using linear least square;Weight mould
Type is established by following formula: y '=ω1*y1+ω2*y2+ω3*y3;Wherein, y ' is standard evaluation score;ω1For fuzzy distortion
Spend linear weight;ω2For the linear weight of noise distortion degree;ω3For the linear weight of the blocking artifact distortion factor;y1It is lost to be fuzzy
True degree;y2For noise distortion degree;y3For the blocking artifact distortion factor.
Second aspect, the embodiment of the present invention also provide a kind of image quality evaluation device, comprising: region unit division module,
For target image to be divided into the multiple regions block of default size;Sharpness value computing module, for calculating multiple regions block
Sharpness value obtains calculated result;Target area extraction module, for extracting target from multiple regions block according to calculated result
Region;Characteristics extraction module, for extracting the characteristic value of target area;Quality evaluation determining module, for according to target area
The characteristic value in domain determines the quality evaluation parameter of target image.
The third aspect, the embodiment of the present invention also provide a kind of electronic equipment, including processor and memory, memory storage
There are the computer executable instructions that can be executed by processor, processor executes computer executable instructions to realize above-mentioned image
The step of quality evaluating method.
Fourth aspect, the embodiment of the present invention also provide a kind of computer readable storage medium, computer readable storage medium
Computer executable instructions are stored with, when being called and being executed by processor, computer is executable to be referred to computer executable instructions
The step of order promotes processor to realize above-mentioned image quality evaluating method.
The embodiment of the present invention bring it is following the utility model has the advantages that
Image quality evaluating method, device and electronic equipment provided in an embodiment of the present invention, target image are divided into more
After a region unit, target area is extracted according to the sharpness value of region unit, characteristic value is extracted from target area and determines target figure
The quality evaluation parameter of picture.The application difficulty that image quality evaluation can be mitigated, increases the accuracy of image quality evaluation.
Other feature and advantage of the disclosure will illustrate in the following description, alternatively, Partial Feature and advantage can be with
Deduce from specification or unambiguously determine, or by implement the disclosure above-mentioned technology it can be learnt that.
To enable the above objects, features, and advantages of the disclosure to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the present invention, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of flow chart of image quality evaluating method provided in an embodiment of the present invention;
Fig. 2 is the flow chart of another image quality evaluating method provided in an embodiment of the present invention;
Fig. 3 is a kind of schematic diagram of the method for building up of weighted model provided in an embodiment of the present invention;
Fig. 4 is a kind of schematic diagram of noise distortion degree calculating process provided in an embodiment of the present invention;
Fig. 5 is a kind of schematic diagram of the establishment process of weighted model provided in an embodiment of the present invention;
Fig. 6 is a kind of structural schematic diagram of image quality evaluation device provided in an embodiment of the present invention;
Fig. 7 is a kind of structural schematic diagram of terminal device provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with attached drawing to the present invention
Technical solution be clearly and completely described, it is clear that described embodiments are some of the embodiments of the present invention, rather than
Whole embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creative work premise
Under every other embodiment obtained, shall fall within the protection scope of the present invention.
Currently, image quality evaluation is divided into subjective assessment and objectively evaluates two kinds.Wherein, subjective assessment refers to direct use
Human eye observes assessment picture quality, not only time-consuming but also with many uncontrollable factors.It objectively evaluates based on computation model pair
Image is evaluated, and is objectively evaluated and is broadly divided into three classes: full reference image quality appraisement, half reference image quality appraisement, without ginseng
Examine image quality evaluation.Full reference picture quality needs whole priori knowledges with original image, requires in practical application severe
It carves.Half reference image quality appraisement reduces application requirement, but also needs original image portion feature.Non-reference picture quality is commented
Valence does not need original image auxiliary, but algorithm is not mature enough at present, and accuracy is to be improved.Non-reference picture quality appraisement master
It is divided into tailored version image quality evaluation algorithm and universal image quality evaluation algorithm.Tailored version evaluation algorithms are for specific
Type of distortion (such as fuzzy, noise, blocking artifact etc.) image carries out quality evaluation, in practical applications, tailored version evaluation algorithms
Different type of distortion can not be adapted to, using being very limited;Although and universal quality evaluation algorithm can be directly to image
Quality is evaluated, but its accuracy need to be improved.Based on this, a kind of image quality evaluation side provided in an embodiment of the present invention
Method, device and electronic equipment, the technology are applied to technical field of image processing, specifically can be adapted for image sharpness region and mention
It takes, image quality evaluation, the fields such as video quality evaluation, is related to image fuzzy detection, edge extracting, Laplace transform, shape
State processing, machine learning etc. technology.
For convenient for understanding the present embodiment, first to a kind of image quality evaluation side disclosed in the embodiment of the present invention
Method describes in detail, as shown in Figure 1, this method comprises the following steps:
Target image is divided into the multiple regions block of default size by step S102.
Target image is exactly the image of pending quality image evaluation, and the type of the image, format and size are unrestricted.
The size of region unit is consistent, and is default size.Default size is determined according to the resolution ratio of target image.Default size compared with
Greatly, then operand is larger, and accuracy is higher, time-consuming more;Default size is smaller, then operand is smaller, and accuracy is lower, time-consuming
It is less.Divide the multiple regions block for referring to and a target image being divided into impartial default size.
Step S104 calculates the sharpness value of multiple regions block, obtains calculated result.
Acutance is also clarity, is an index for reflecting plane of delineation clarity and the sharp keen degree in image border.It calculates
It as a result is exactly the sharpness value of each region unit.
Step S106 extracts target area from multiple regions block according to calculated result.
According to the size of calculated result, a part of region unit is extracted from all areas block that target image divides, it is above-mentioned
A part of region unit is exactly target area.The purpose that target area is extracted is to extract part more visible in target image.
Step S108 extracts the characteristic value of target area.
Characteristic value can be understood as the change degree that target area is carried out to the fuzzy front and back of image.
Step S110 determines the quality evaluation parameter of target image according to the characteristic value of target area.
Quality evaluation parameter is used to reflect the quality of target image.The target that quality evaluation parameter is extracted according to target image
The characteristic value in region determines.
A kind of image quality evaluating method provided in an embodiment of the present invention, after target image is divided into multiple regions block,
Target area is extracted according to the sharpness value of region unit, extract characteristic value from target area and determines the quality evaluation of target image
Parameter.The application difficulty that image quality evaluation can be mitigated, increases the accuracy of image quality evaluation.
The embodiment of the present invention also provides another image quality evaluating method;This method is on the basis of above-described embodiment method
Upper realization;The description of this method emphasis extracts the specific implementation of target area according to calculated result from multiple regions block.
As shown in Fig. 2, this method comprises the following steps:
Target image is divided into the multiple regions block of default size by step S202.
Default size is preset according to the resolution ratio of target image, and in general, the value range of default size exists
[32*32,160*160], preferred value 96*96, unit are pixel * pixel.Region unit is generally small rectangular area, as target
The fundamental unit of extracted region.
Step S204 calculates the sharpness value of multiple regions block, obtains calculated result.
Sharpness value generally passes through the sharpness value of each pixel in the block of zoning, and is summed to obtain, specifically,
For each region unit, the sharpness value of each pixel in current region block is calculated according to following formula:
By the sum of the sharpness value of pixel each in current region block, it is determined as current region block sharpness value;
Wherein, I (i, j) is the pixel coordinate in current region block;σ (i, j) is the sharpness value of pixel I (i, j);
ωk.lIt is the discrete two-dimensional Gaussian function coefficient using k and l as parameter;Any integer value of the k between value range-K to K;L is
Any integer value between value range-L to L;K and L is preset fixed value.
σ (i, j) can be understood as the variance yields of pixel I (i, j), which indicates each pixel relative to image averaging
The dispersion of gray scale can indicate the sharpness value of the pixel.Acutance shows that more greatly details differentiation is more obvious, more similar to variance
Clearly.But for the region of not details, such as simple white, and clearly.If piece image is pure color entirely, do not have yet
The meaning for having judgement fuzzy.But the accuracy of above-mentioned image quality evaluating method is nor affected on, because of above-mentioned image quality evaluation
Method itself is exactly clear image to such case judgement, meets the fact.
It should be noted that K and L can be set to 3, i.e. K=L=3.ω={ ωk.l| k=-3, -2, -1,0,1,2,
3, l=-3, -2, -1,0,1,2,3, }.
Step S206 is ranked up multiple regions block, obtains ranking results according to the sequence that sharpness value is descending.
The calculating to sharpness value is carried out to all areas block according to step S204, and has obtained calculated result.By sharpness value
Calculated result sort from large to small, all areas block is ranked up and is obtained with corresponding ranking results.
Step S208 extracts the region unit of continuous specified quantity from first region BOB(beginning of block) in ranking results.
In general, region unit is extracted into the biggish region of continuous sharpness value according to the size descending sort of sharpness value
Block.The quantity of extraction is usually the region unit of the biggish preceding 10%-60% of sharpness value, preferred value 30%, i.e. extraction sharpness value
The biggish region unit preceding 30% is optimal.
The region unit extracted is determined as target area by step S210.
Using the biggish region unit of the sharpness value of extraction as target area.For example, target image is divided into multiple small squares
The region unit of shape indicates the target area block extracted with circle, regards target area.The numerical value of each region unit is exactly
The calculated result of sharpness value.The local clear area that can be very good in screening target image, that is, target area are done so,
It lays a good foundation further to promote the accuracy of image quality evaluating method.
Step S212 extracts the characteristic value of target area.
The characteristic value of target area is extracted, specifically, refers to and Gaussian Blur is carried out to target area, then calculate Gauss
The change degree of fuzzy front and back, using change degree as the characteristic value of the target area.Gaussian Blur is also Gaussian smoothing, for reducing
Picture noise and reduction level of detail.
Specifically, the characteristic value of target area can be extracted by following steps:
(1) Fuzzy Processing is carried out to target area, the target area after obtaining Fuzzy Processing.
Fuzzy Processing is exactly above-mentioned Gaussian Blur.
(2) characteristic value of target area is calculated by following formula:
Wherein, Δ t is the characteristic value of target area;N is the number of target area;I is value range 0 to appointing between N
Meaning integer value;F is that preset change degree calculates function;PiFor target area, P 'iFor the target area after Fuzzy Processing.
Wherein, the change degree that change degree calculates that function F is fuzzy fuzzy algorithmic approach again calculates function, calculates in two steps: first,
In PiWith P 'iIt is interior, calculate separately the second order difference value of pixel abscissa and ordinate direction;Second, calculate PiWith P 'iIn it is each
The difference of pixel second differnce, by the difference in region and as change degree.
Step S214 determines the quality evaluation parameter of target image according to the characteristic value of target area.
According to Gaussian Blur principle it is found that picture quality and characteristic value Δ t correlation, which can be by linear mould
Type description.Characteristic value Δ t is calculated by the way of in step S212, passes through the linear mould between picture quality and characteristic value Δ t
Type can obtain picture quality.Specifically, can be executed by following steps: the characteristic value input of target area is preset
In weighted model, the quality evaluation parameter of target image is exported;Weighted model is imitated according to the fuzzy distortion factor, noise distortion degree and block
The distortion factor is answered to establish.Weighted model is exactly the linear model between above-mentioned picture quality and characteristic value Δ t, and quality evaluation parameter is used
Carry out the picture quality of evaluation goal image.
The method for building up of weighted model in step S214, as shown in figure 3, this method comprises the following steps:
Step S502 determines the standard evaluation score of training image and training image based on preset training set.
Training pattern needs training set, here using University of Texas LIVE image library as training set, in training set
Comprising different type distorted image, while also comprising the corresponding subjective assessment point of image.Distorted image is exactly training image, and is led
See the standard evaluation score that evaluation point is exactly training image.
Step S504 calculates the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor of training image.
The fuzzy distortion factor of training image is calculated using the blurred picture quality evaluation algorithm based on human eye characteristic.Use public affairs
Open the noise distortion degree that Noise Algorithm calculates training image.It is distorted using the blocking artifact that open blocking artifact algorithm calculates training image
Degree.
Wherein, the schematic diagram of the feature calculation of noise distortion degree a kind of noise distortion degree calculating process shown in Figure 4,
In the picture, only noise and marginal point belongs to high-frequency information.In high-frequency information, removal marginal information can obtain noise
Information.
Firstly, it is necessary to carry out Sobel edge detection and binary conversion treatment to training image, can by following formula into
Row:
I (i, j) is the pixel value at image coordinate (i, j), and G (i, j) is the gradient magnitude at image coordinate (i, j), Gx
(i, j) is the abscissa component of G (i, j), Gy(i, j) is the ordinate component of G (i, j), and wherein * represents convolution.Use threshold value
By image binaryzation, (i.e. G (i, j) > K is white, other are all black;Wherein K takes according to the statistics with histogram of the every gradient value of image
85% partition value before statistical sample).
Since bianry image (being denoted as Gray) both contains marginal information, also contain noise information, needs to utilize image expansion
Marginal information is removed with etching operation.In addition to independent point in grayscale image, remaining is marginal information entirely, is traversed in grayscale image
Point chooses independent point (i.e. only oneself being the point of white in 8 fields), and remaining white point is denoted as marginal point.It calculates
The Laplacian operator coefficient of noise information, as the high-frequency characteristic of image, noise is the point of non-edge in original image.
It is calculated by the following formula noise distortion degree:Wherein, n is the number of pixels of training image, E
For the noise distortion degree of training image, V is Laplacian operator.
Since blocking artifact is mainly caused by the quantization error after discrete cosine transform, dct basis our unit is 8*
8 block of pixels.Change degree by calculating block boundary can measure the blocking artifact distortion factor.It is small that training image is divided into 8*8 first
Block, calculates each piece of level, the change degree and zero-crossing rate in vertical direction block edge average variability, block with block edge.Root
Upper parameter carries out exponential model training using image data base accordingly, and carries out the blocking artifact distortion factor using the model after training and comment
Valence.
The fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor is normalized in step S506.
Since the unit of every kind of algorithm is different, the scoring to each algorithm is needed to be normalized.Such i-th width instruction
Practice image Ii, one group of three distortion factor can be corresponded to, that is, the fuzzy distortion factor, noise distortion degree and blocking artifact distortion after normalizing
Degree.Wherein, y1iThe fuzzy distortion factor after being denoted as normalization, y2iNoise distortion degree after being denoted as normalization, y3iAfter being denoted as normalization
The blocking artifact distortion factor.
Step S508 is adjusted according to the standard evaluation score of training image and is obscured the distortion factor, noise distortion in weighted model
Spend weight corresponding with the blocking artifact distortion factor.
The standard evaluation score of each training image is had been presented in training set.According to the standard evaluation score, with
And the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor after normalization, so that it may carry out model training, calculate fuzzy lose
True degree, noise distortion degree and the corresponding weight of the blocking artifact distortion factor.It can be executed by following steps:
According to the standard evaluation score of training image, the fuzzy distortion factor, noise distortion are calculated using linear least square
The weight of degree and the blocking artifact distortion factor;
Weighted model is established by following formula:
Y '=ω1*y1+ω2*y2+ω3*y3;
Wherein, y ' is standard evaluation score;ω1To obscure the linear weight of the distortion factor;ω2Linear for noise distortion degree
Weight;ω3For the linear weight of the blocking artifact distortion factor;y1To obscure the distortion factor;y2For noise distortion degree;y3For blocking artifact distortion
Degree.
Step S510 continues to execute the step of determining training image based on preset training set, until according to weight calculation
The error of obtained evaluation score and standard evaluation score within a preset range, obtains final weighted model.
Foundation for weighted model, generally requires one preset range of setting, which can pass through percentage
It indicates, generally between 5-20%, preferred value 10%.The evaluation score and standard evaluation score obtained according to weight calculation
Error within 10%, then retain corresponding weight, obtain final weighted model.
The establishment process of model may refer to a kind of schematic diagram of the establishment process of weighted model shown in fig. 5, firstly, from
Training set obtains a training image and its standard evaluation score, obtains the fuzzy of the training image respectively with different algorithms
The distortion factor, noise distortion degree and the blocking artifact distortion factor, and be normalized respectively, by the fuzzy distortion after normalized
Degree, noise distortion degree and the blocking artifact distortion factor are applied in weighted model, export an overall merit, are exactly above-mentioned evaluation point
Number.The evaluation score is made with standard evaluation score and is compared, sees the error of the two whether in default range.If error
In default range, then continue to train using next width training image;If it was not then needing to modify weight, to guarantee to repair
Error after changing is in default range.
In aforesaid way, by obtaining ranking results to the descending sequence of sharpness value, and choose sharpness value it is biggish
Region unit in preset quantity is as target area;Using the change degree after the Fuzzy Processing of target area as target area
Characteristic value;The quality evaluation parameter of target image is exported by weighted model.Accuracy rate is higher, and screens zoning, drop
Low calculating cost improves the robustness and accuracy of image quality evaluation.
It should be noted that the embodiments are all described in a progressive manner for above-mentioned each method, each embodiment is stressed
Be the difference from other embodiments, the same or similar parts between the embodiments can be referred to each other.
Corresponding to above method embodiment, the embodiment of the invention provides a kind of image quality evaluation devices, such as Fig. 6 institute
Show, which includes:
Region unit division module 81, for target image to be divided into the multiple regions block of default size;
Sharpness value computing module 82 obtains calculated result for calculating the sharpness value of multiple regions block;
Target area extraction module 83, for extracting target area from multiple regions block according to calculated result;
Characteristics extraction module 84, for extracting the characteristic value of target area;
Quality evaluation determining module 85 determines the quality evaluation ginseng of target image for the characteristic value according to target area
Number.
A kind of image quality evaluation device provided in an embodiment of the present invention, after target image is divided into multiple regions block,
Target area is extracted according to the sharpness value of region unit, extract characteristic value from target area and determines the quality evaluation of target image
Parameter.The application difficulty that image quality evaluation can be mitigated, increases the accuracy of image quality evaluation.
In some embodiments, sharpness value computing module is used for: being directed to each region unit, is worked as according to the calculating of following formula
The sharpness value of each pixel in forefoot area block:By pixel each in current region block
The sum of sharpness value is determined as current region block sharpness value;Wherein, I (i, j) is the pixel coordinate in current region block;σ (i,
J) sharpness value for being pixel I (i, j);ωk.lIt is the discrete two-dimensional Gaussian function coefficient using k and l as parameter;K is value model
Enclose any integer value between-K to K;Any integer value of the l between value range-L to L;K and L is preset fixation
Value.
In some embodiments, target area extraction module is used for: according to the sequence that sharpness value is descending, to multiple
Region unit is ranked up, and obtains ranking results;From first region BOB(beginning of block) in ranking results, continuous specified quantity is extracted
Region unit;The region unit extracted is determined as target area.
In some embodiments, characteristics extraction module is used for: being carried out Fuzzy Processing to target area, is obtained fuzzy place
Target area after reason;The characteristic value of target area is calculated by following formula:Wherein, Δ t is target
The characteristic value in region;N is the number of target area;I is value range 0 to any integer value between N;F is preset variation
Degree calculates function;PiFor target area, P 'iFor the target area after Fuzzy Processing.
In some embodiments, quality evaluation determining module is used for: the characteristic value input of target area is preset
In weighted model, the quality evaluation parameter of target image is exported;Weighted model is imitated according to the fuzzy distortion factor, noise distortion degree and block
The distortion factor is answered to establish.
In some embodiments, weighted model is established by following steps: training image is determined based on preset training set,
And the standard evaluation score of training image;Calculate the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor of training image;
The fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor are normalized;According to the evaluation point of the standard of training image
Number adjusts and obscures the distortion factor, noise distortion degree and the corresponding weight of the blocking artifact distortion factor in weighted model;It continues to execute based on pre-
If training set the step of determining training image, until the mistake of the evaluation score and standard evaluation score obtained according to weight calculation
Difference within a preset range, obtains final weighted model.
In some embodiments, weighted model is used for: according to the standard evaluation score of training image, using linear minimum
Square law calculates the fuzzy distortion factor, the weight of noise distortion degree and the blocking artifact distortion factor;Weighted model is established by following formula:
Y '=ω1*y1+ω2*y2+ω3*y3;Wherein, y ' is standard evaluation score;ω1To obscure the linear weight of the distortion factor;ω2To make an uproar
The linear weight of sound distortion degree;ω3For the linear weight of the blocking artifact distortion factor;y1To obscure the distortion factor;y2For noise distortion degree;
y3For the blocking artifact distortion factor.
Image quality evaluation device provided in an embodiment of the present invention, with image quality evaluation dress side provided by the above embodiment
Method technical characteristic having the same reaches identical technical effect so also can solve identical technical problem.
The embodiment of the invention also provides a kind of electronic equipment, for running above-mentioned image quality evaluating method;Referring to Fig. 7
Shown, which includes memory 100 and processor 101, wherein memory 100 is calculated for storing one or more
Machine instruction, one or more computer instruction is executed by processor 101, to realize above-mentioned image quality evaluating method.
Further, electronic equipment shown in Fig. 7 further includes bus 102 and communication interface 103, and processor 101, communication connect
Mouth 103 and memory 100 are connected by bus 102.
Wherein, memory 100 may include high-speed random access memory (RAM, Random Access Memory),
It may further include non-labile memory (non-volatile memory), for example, at least a magnetic disk storage.By extremely
A few communication interface 103 (can be wired or wireless) is realized logical between the system network element and at least one other network element
Letter connection, can be used internet, wide area network, local network, Metropolitan Area Network (MAN) etc..Bus 102 can be isa bus, pci bus or
Eisa bus etc..Bus can be divided into address bus, data/address bus, control bus etc..For convenient for indicating, only with one in Fig. 7
Four-headed arrow indicates, it is not intended that an only bus or a type of bus.
Processor 101 may be a kind of IC chip, the processing capacity with signal.It is above-mentioned during realization
Each step of method can be completed by the integrated logic circuit of the hardware in processor 101 or the instruction of software form.On
The processor 101 stated can be general processor, including central processing unit (Central Processing Unit, abbreviation
CPU), network processing unit (Network Processor, abbreviation NP) etc.;It can also be digital signal processor (Digital
Signal Processor, abbreviation DSP), specific integrated circuit (Application Specific Integrated
Circuit, abbreviation ASIC), field programmable gate array (Field-Programmable Gate Array, abbreviation FPGA) or
Person other programmable logic device, discrete gate or transistor logic, discrete hardware components.It may be implemented or execute sheet
Disclosed each method, step and logic diagram in inventive embodiments.General processor can be microprocessor or the processing
Device is also possible to any conventional processor etc..The step of method in conjunction with disclosed in the embodiment of the present invention, can be embodied directly in
Hardware decoding processor executes completion, or in decoding processor hardware and software module combination execute completion.Software mould
Block can be located at random access memory, flash memory, read-only memory, programmable read only memory or electrically erasable programmable storage
In the storage medium of this fields such as device, register maturation.The storage medium is located at memory 100, and processor 101 reads memory
Information in 100, in conjunction with its hardware complete previous embodiment method the step of.
The embodiment of the invention also provides a kind of computer readable storage medium, which has
Computer executable instructions, when being called and being executed by processor, computer executable instructions promote the computer executable instructions
Processor is set to realize above-mentioned image quality evaluating method, specific implementation can be found in embodiment of the method, and details are not described herein.
The computer program product of image quality evaluating method, device provided by the embodiment of the present invention and electronic equipment,
Computer readable storage medium including storing program code, the instruction that program code includes can be used for executing previous methods reality
The method in example is applied, specific implementation can be found in embodiment of the method, and details are not described herein.
It is apparent to those skilled in the art that for convenience and simplicity of description, the device of foregoing description
And/or the specific work process of electronic equipment, it can refer to corresponding processes in the foregoing method embodiment, details are not described herein.
Finally, it should be noted that above embodiments, only a specific embodiment of the invention, to illustrate skill of the invention
Art scheme, rather than its limitations, scope of protection of the present invention is not limited thereto, although with reference to the foregoing embodiments to the present invention into
Go detailed description, those skilled in the art should understand that: anyone skilled in the art is at this
It invents in the technical scope disclosed, can still modify or can be thought easily to technical solution documented by previous embodiment
To variation or equivalent replacement of some of the technical features;And these modifications, variation or replacement, do not make corresponding
The essence of technical solution is detached from the spirit and scope of technical solution of the embodiment of the present invention, should all cover in protection scope of the present invention
Within.Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. a kind of image quality evaluating method characterized by comprising
Target image is divided into the multiple regions block of default size;
The sharpness value for calculating multiple region units, obtains calculated result;
According to the calculated result, target area is extracted from multiple region units;
Extract the characteristic value of the target area;
According to the characteristic value of the target area, the quality evaluation parameter of the target image is determined.
2. being calculated the method according to claim 1, wherein calculating the sharpness value of multiple region units
As a result the step of, comprising:
For each region unit, the sharpness value of each pixel in current region block is calculated according to following formula:
By the sum of the sharpness value of each pixel in the current region block, it is determined as the current region block sharpness value;
Wherein, I (i, j) is the pixel coordinate in current region block;σ (i, j) is the sharpness value of pixel I (i, j);ωk.lIt is
Using k and l as the discrete two-dimensional Gaussian function coefficient of parameter;Any integer value of the k between value range-K to K;L is value model
Enclose any integer value between-L to L;K and L is preset fixed value.
3. the method according to claim 1, wherein according to the calculated result, from multiple region units
The step of extracting target area, comprising:
According to the sequence that sharpness value is descending, multiple region units are ranked up, ranking results are obtained;
From first region BOB(beginning of block) in the ranking results, the region unit of continuous specified quantity is extracted;
The region unit extracted is determined as target area.
4. the method according to claim 1, wherein the step of extracting the characteristic value of the target area, comprising:
Fuzzy Processing is carried out to the target area, the target area after obtaining Fuzzy Processing;
The characteristic value of the target area is calculated by following formula:
Wherein, Δ t is the characteristic value of the target area;N is the number of the target area;I is value range 0 between N
Any integer value;F is that preset change degree calculates function;PiFor the target area, P 'iFor the mesh after Fuzzy Processing
Mark region.
5. the method according to claim 1, wherein determining the mesh according to the characteristic value of the target area
The step of quality evaluation parameter of logo image, comprising:
The characteristic value of the target area is inputted in preset weighted model, the quality evaluation of the target image is exported
Parameter;The weighted model is established according to the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor.
6. according to the method described in claim 5, it is characterized in that, the weighted model is established by following steps:
The standard evaluation score of training image and the training image is determined based on preset training set;
Calculate the fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor of the training image;
The fuzzy distortion factor, noise distortion degree and the blocking artifact distortion factor are normalized;
According to the standard evaluation score of the training image, adjust obscured described in weighted model the distortion factor, noise distortion degree and
The corresponding weight of the blocking artifact distortion factor;
The step of training image is determined based on preset training set is continued to execute, until the evaluation obtained according to the weight calculation
The error of score and the standard evaluation score within a preset range, obtains final weighted model.
7. according to the method described in claim 6, it is characterized in that, being adjusted according to the standard evaluation score of the training image
The step of distortion factor, noise distortion degree and the corresponding weight of the blocking artifact distortion factor are obscured described in initial model, comprising:
According to the standard evaluation score of the training image, the fuzzy distortion factor, noise are calculated using linear least square
The weight of the distortion factor and the blocking artifact distortion factor;
The weighted model is established by following formula:
Y '=ω1*y1+ω2*y2+ω3*y3;
Wherein, y ' is the standard evaluation score;ω1To obscure the linear weight of the distortion factor;ω2Linear for noise distortion degree
Weight;ω3For the linear weight of the blocking artifact distortion factor;y1For the fuzzy distortion factor;y2For the noise distortion degree;y3For institute
State the blocking artifact distortion factor.
8. a kind of image quality evaluation device characterized by comprising
Region unit division module, for target image to be divided into the multiple regions block of default size;
Sharpness value computing module obtains calculated result for calculating the sharpness value of multiple region units;
Target area extraction module, for extracting target area from multiple region units according to the calculated result;
Characteristics extraction module, for extracting the characteristic value of the target area;
Quality evaluation determining module determines the quality evaluation of the target image for the characteristic value according to the target area
Parameter.
9. a kind of electronic equipment, which is characterized in that including processor and memory, the memory is stored with can be by the place
The computer executable instructions that device executes are managed, the processor executes the computer executable instructions to realize claim 1
The step of to 7 described in any item image quality evaluating methods.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer can
It executes instruction, when being called and being executed by processor, the computer executable instructions promote the computer executable instructions
Processor realizes the step of claim 1 to 7 described in any item image quality evaluating methods.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910674392.2A CN110378893B (en) | 2019-07-24 | 2019-07-24 | Image quality evaluation method and device and electronic equipment |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910674392.2A CN110378893B (en) | 2019-07-24 | 2019-07-24 | Image quality evaluation method and device and electronic equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110378893A true CN110378893A (en) | 2019-10-25 |
CN110378893B CN110378893B (en) | 2021-11-16 |
Family
ID=68255757
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910674392.2A Active CN110378893B (en) | 2019-07-24 | 2019-07-24 | Image quality evaluation method and device and electronic equipment |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110378893B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275681A (en) * | 2020-01-19 | 2020-06-12 | 浙江大华技术股份有限公司 | Picture quality determination method and device, storage medium and electronic device |
CN111539914A (en) * | 2020-03-24 | 2020-08-14 | 上海交通大学 | Mobile phone photo quality comparison and evaluation method, system and terminal |
CN111899243A (en) * | 2020-07-28 | 2020-11-06 | 阳光保险集团股份有限公司 | Image definition evaluation method and device and computer readable storage medium |
CN112102309A (en) * | 2020-09-27 | 2020-12-18 | 中国建设银行股份有限公司 | Method, device and equipment for determining image quality evaluation result |
CN115995050A (en) * | 2023-03-22 | 2023-04-21 | 山东汇蓝环保科技有限公司 | Environment monitoring data processing system based on aerial photography |
Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030053544A1 (en) * | 2001-09-18 | 2003-03-20 | Tomoko Yasunari | Method and apparatus for motion vector detection and medium storing method program directed to the same |
CN102609939A (en) * | 2012-01-16 | 2012-07-25 | 北京航空航天大学 | TFDS (Train Coach Machine Vision Detection System) image quality evaluation method and system |
CN102622737A (en) * | 2012-02-29 | 2012-08-01 | 宁波大学 | Microscopic image fusion method based on region growth |
CN102663747A (en) * | 2012-03-23 | 2012-09-12 | 宁波大学 | Stereo image objectivity quality evaluation method based on visual perception |
CN102740114A (en) * | 2012-07-16 | 2012-10-17 | 公安部第三研究所 | Non-parameter evaluation method for subjective quality of video |
CN103617617A (en) * | 2013-12-05 | 2014-03-05 | 淮海工学院 | Underwater image quality evaluating and measuring method based on power spectrum description |
CN103632369A (en) * | 2013-12-05 | 2014-03-12 | 淮海工学院 | Method for universally detecting quality of non-reference underwater images on basis of combination of block average definition |
US8692884B2 (en) * | 2010-09-17 | 2014-04-08 | Gish Technology, Inc. | Apparatus and method for assessing visual acuity |
CN103810702A (en) * | 2014-01-17 | 2014-05-21 | 浙江大学 | Blocking noise variance estimation based image accumulation signal to noise ratio calculation method |
CN104050678A (en) * | 2014-07-01 | 2014-09-17 | 淮海工学院 | Underwater monitoring color image quality measurement method |
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 |
CN104637064A (en) * | 2015-02-28 | 2015-05-20 | 中国科学院光电技术研究所 | Defocus blurred image definition detection method based on edge intensity weight |
CN105744256A (en) * | 2016-03-31 | 2016-07-06 | 天津大学 | Three-dimensional image quality objective evaluation method based on graph-based visual saliency |
US20160327625A1 (en) * | 2013-12-20 | 2016-11-10 | Irccs Centro Neurolesi "Bonino-Pulejo" | Universal phantom structure for quality inspections both on computerized tomography and on magnetic resonance tomography |
CN106791800A (en) * | 2016-11-16 | 2017-05-31 | 深圳六滴科技有限公司 | The quality diagnosis method and device of panoramic video |
CN106856002A (en) * | 2016-11-22 | 2017-06-16 | 上海大学 | A kind of unmanned plane shooting image quality evaluating method |
CN109447952A (en) * | 2018-10-10 | 2019-03-08 | 嘉兴学院 | A kind of half reference image quality evaluation method based on Gabor differential box weighting dimension |
-
2019
- 2019-07-24 CN CN201910674392.2A patent/CN110378893B/en active Active
Patent Citations (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030053544A1 (en) * | 2001-09-18 | 2003-03-20 | Tomoko Yasunari | Method and apparatus for motion vector detection and medium storing method program directed to the same |
US8692884B2 (en) * | 2010-09-17 | 2014-04-08 | Gish Technology, Inc. | Apparatus and method for assessing visual acuity |
CN102609939A (en) * | 2012-01-16 | 2012-07-25 | 北京航空航天大学 | TFDS (Train Coach Machine Vision Detection System) image quality evaluation method and system |
CN102622737A (en) * | 2012-02-29 | 2012-08-01 | 宁波大学 | Microscopic image fusion method based on region growth |
CN102663747A (en) * | 2012-03-23 | 2012-09-12 | 宁波大学 | Stereo image objectivity quality evaluation method based on visual perception |
CN102740114A (en) * | 2012-07-16 | 2012-10-17 | 公安部第三研究所 | Non-parameter evaluation method for subjective quality of video |
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 |
CN103617617A (en) * | 2013-12-05 | 2014-03-05 | 淮海工学院 | Underwater image quality evaluating and measuring method based on power spectrum description |
CN103632369A (en) * | 2013-12-05 | 2014-03-12 | 淮海工学院 | Method for universally detecting quality of non-reference underwater images on basis of combination of block average definition |
US20160327625A1 (en) * | 2013-12-20 | 2016-11-10 | Irccs Centro Neurolesi "Bonino-Pulejo" | Universal phantom structure for quality inspections both on computerized tomography and on magnetic resonance tomography |
CN103810702A (en) * | 2014-01-17 | 2014-05-21 | 浙江大学 | Blocking noise variance estimation based image accumulation signal to noise ratio calculation method |
CN104050678A (en) * | 2014-07-01 | 2014-09-17 | 淮海工学院 | Underwater monitoring color image quality measurement method |
CN104637064A (en) * | 2015-02-28 | 2015-05-20 | 中国科学院光电技术研究所 | Defocus blurred image definition detection method based on edge intensity weight |
CN105744256A (en) * | 2016-03-31 | 2016-07-06 | 天津大学 | Three-dimensional image quality objective evaluation method based on graph-based visual saliency |
CN106791800A (en) * | 2016-11-16 | 2017-05-31 | 深圳六滴科技有限公司 | The quality diagnosis method and device of panoramic video |
CN106856002A (en) * | 2016-11-22 | 2017-06-16 | 上海大学 | A kind of unmanned plane shooting image quality evaluating method |
CN109447952A (en) * | 2018-10-10 | 2019-03-08 | 嘉兴学院 | A kind of half reference image quality evaluation method based on Gabor differential box weighting dimension |
Non-Patent Citations (2)
Title |
---|
GHISLAIN TAKAM TCHENDJOU等: ""Fuzzy logic modeling for objective image quality assessment"", 《2016 CONFERENCE ON DESIGN AND ARCHITECTURES FOR SIGNAL AND IMAGE PROCESSING (DASIP)》 * |
徐少平等: ""信息量加权的梯度显著度图像质量评价"", 《中国国象图形学报》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111275681A (en) * | 2020-01-19 | 2020-06-12 | 浙江大华技术股份有限公司 | Picture quality determination method and device, storage medium and electronic device |
CN111275681B (en) * | 2020-01-19 | 2023-09-01 | 浙江大华技术股份有限公司 | Picture quality determining method and device, storage medium and electronic device |
CN111539914A (en) * | 2020-03-24 | 2020-08-14 | 上海交通大学 | Mobile phone photo quality comparison and evaluation method, system and terminal |
CN111899243A (en) * | 2020-07-28 | 2020-11-06 | 阳光保险集团股份有限公司 | Image definition evaluation method and device and computer readable storage medium |
CN112102309A (en) * | 2020-09-27 | 2020-12-18 | 中国建设银行股份有限公司 | Method, device and equipment for determining image quality evaluation result |
CN115995050A (en) * | 2023-03-22 | 2023-04-21 | 山东汇蓝环保科技有限公司 | Environment monitoring data processing system based on aerial photography |
Also Published As
Publication number | Publication date |
---|---|
CN110378893B (en) | 2021-11-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110378893A (en) | Image quality evaluating method, device and electronic equipment | |
Ahuja | A transform for multiscale image segmentation by integrated edge and region detection | |
Annangi et al. | A region based active contour method for x-ray lung segmentation using prior shape and low level features | |
US20170365046A1 (en) | Algorithm and device for image processing | |
WO2017121018A1 (en) | Method and apparatus for processing two-dimensional code image, and terminal and storage medium | |
Xiaofeng et al. | Discriminatively trained sparse code gradients for contour detection | |
CN108921800A (en) | Non-local mean denoising method based on form adaptive search window | |
Bora et al. | A novel approach towards clustering based image segmentation | |
CN111241927A (en) | Cascading type face image optimization method, system and equipment and readable storage medium | |
CN107633491A (en) | A kind of area image Enhancement Method and storage medium based on target detection | |
CN110717922A (en) | Image definition evaluation method and device | |
CN113592776A (en) | Image processing method and device, electronic device and storage medium | |
CN102222328A (en) | Edge-preserving self-adaptive weighted filtering method for natural scene images | |
CN111612741A (en) | Accurate non-reference image quality evaluation method based on distortion recognition | |
CN109711268A (en) | A kind of facial image screening technique and equipment | |
CN104240259B (en) | High photographing instrument voucher intelligence cutting edge correction system and method based on contours segmentation | |
CN112184672A (en) | No-reference image quality evaluation method and system | |
CN115578284A (en) | Multi-scene image enhancement method and system | |
Ma et al. | Efficient saliency analysis based on wavelet transform and entropy theory | |
CN110569859A (en) | Color feature extraction method for clothing image | |
CN106683063A (en) | Method and device of image denoising | |
Jia et al. | Fabric defect inspection based on lattice segmentation and lattice templates | |
Tan et al. | Image haze removal based on superpixels and Markov random field | |
Shukla et al. | Adaptive fractional masks and super resolution based approach for image enhancement | |
WO2017128646A1 (en) | Image processing method and device |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |