CN107203771A - Database building method - Google Patents
Database building method Download PDFInfo
- Publication number
- CN107203771A CN107203771A CN201710484335.9A CN201710484335A CN107203771A CN 107203771 A CN107203771 A CN 107203771A CN 201710484335 A CN201710484335 A CN 201710484335A CN 107203771 A CN107203771 A CN 107203771A
- Authority
- CN
- China
- Prior art keywords
- aesthetic
- feature
- image
- factors
- described image
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/148—Wavelet transforms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Software Systems (AREA)
- Computational Mathematics (AREA)
- Multimedia (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Algebra (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to technical field of image processing, in particular to a kind of database building method.The database building method that the present invention is provided, scored according to the machine of aesthetic factors, the artificial scoring of the scoring weight of each mark person and each mark person to aesthetic factors, calculate the overall score of the aesthetic factors of described image sample, the judged result inputted according to multiple mark persons, calculate the scoring of the overall aesthetic quality of described image sample, and according to first threshold, Second Threshold, 3rd threshold value, image pattern is labeled as high/low composition aesthetic feeling quality image by the 4th threshold value and the 5th threshold value, high/low brightness aesthetic feeling quality image, high/low color aesthetic feeling quality image, high/low depth of field aesthetic feeling quality image and high/low overall aesthetic quality image, realize the foundation of database.The database that the database building method provided using the present invention is set up can be used for the careful research of image aesthetic feeling quality.
Description
Technical field
The present invention relates to technical field of image processing, in particular to a kind of database building method.
Background technology
In recent decades, analysis and the research of the aesthetic feeling quality of image are increasingly paid close attention to by people.At present still
The database dedicated for the careful research of image aesthetic feeling quality is not proposed.
The content of the invention
In view of this, it is an object of the invention to provide a kind of database building method, to solve the above problems.
To achieve the above object, the present invention provides following technical scheme:
A kind of database building method, the database is used for image aesthetic feeling quality evaluation, and methods described includes:
The multiple image of multi-class, different qualities is obtained, to be used as image pattern;
Calculate described image sample aesthetic factors machine scoring, wherein, the aesthetic factors include composition aesthetic feeling because
Element, brightness aesthetic factors, color aesthetic factors and depth of field aesthetic factors;
Survey is carried out to mark person, the problem of being inputted according to each mark person answer calculates the scoring power of mark person
Weight;
The qualitative of aesthetic factors of described image sample is judged according to each mark person, each mark person is calculated to described image
The artificial scoring of the aesthetic factors of sample;
According to the scoring of the machine of the aesthetic factors of described image sample, the scoring weight of each mark person and each mark person couple
The artificial scoring of the aesthetic factors of described image sample, calculates the overall score of the aesthetic factors of described image sample, including
The overall scores of the composition aesthetic factors of image pattern, the overall score of brightness aesthetic factors, the overall score of color aesthetic factors and
The overall score of depth of field aesthetic factors;
The judged result inputted according to multiple mark persons, calculates the scoring of the overall aesthetic quality of described image sample;
According to the overall score of the composition aesthetic factors of described image sample by image labeling be high/low composition aesthetic feeling quality figure
Picture, according to the overall score of the brightness aesthetic factors of described image sample by image labeling be high/low brightness aesthetic feeling quality image, root
According to described image sample color aesthetic factors overall score by image labeling be high/low color aesthetic feeling quality image, according to institute
Image labeling is high/low depth of field aesthetic feeling quality image and according to the figure by the overall score for stating the depth of field aesthetic factors of image pattern
Image labeling is high/low depth of field aesthetic feeling quality image by the scoring of the overall aesthetic quality of decent, to set up database.
Alternatively, the step of machine for calculating the aesthetic factors of described image sample scores includes:
Calculate the machine scoring of the feature of the aesthetic factors of described image sample;
Calculate the average value of the machine scoring of the feature of the aesthetic factors of described image sample;
When the average value of the machine scoring of the feature of the aesthetic factors of described image sample is less than 0.1, described image sample
The machine scoring of this aesthetic factors is equal to 1;
It is described when the average value of the machine scoring of the feature of the aesthetic factors of described image sample is more than or equal to 0.1
The machine scoring of the aesthetic factors of image pattern is equal to 10 times of average values.
Alternatively, the step of machine of the feature of the aesthetic factors for calculating described image sample scores includes:
Calculate the characteristic value of the feature of the aesthetic factors of described image sample;
According to the characteristic value of the feature of multiple high aesthetic feeling quality images, the high aesthetic feeling quality span of the feature is calculated
And optimal value;
According to the characteristic value of the feature of described image sample and the high aesthetic feeling quality span and optimal value of the feature
Calculate the machine scoring of the feature of described image sample.
Alternatively, according to the characteristic value of the feature of multiple high aesthetic feeling quality images, the high aesthetic feeling quality of the feature is calculated
The step of span and optimal value, includes:
According to the characteristic value of each feature of multiple high aesthetic feeling quality images, each feature histogram is drawn respectively, wherein, it is described
The width of the Nogata post of each feature histogram uses Scott derivation formulas, and the ordinate of each feature histogram represents that this is straight
The picture number of high aesthetic feeling quality image corresponding to square column, abscissa represents the characteristic value of the feature corresponding to the Nogata post;
The extreme difference of the ordinate of each feature histogram is calculated, ordinate in each feature histogram is obtained and is more than or equal to
The abscissa scope of the Nogata post of 0.5 times of extreme difference, as the high aesthetic feeling quality span of each feature, is designated as [Vai,
Vbi];
Obtain the abscissa zone that ordinate in each feature histogram is more than or equal to the Nogata post of 0.9 times of maximum
Intermediate value as the optimal value of each feature, be designated as Vopti。
Alternatively, the high aesthetic feeling quality value model of the characteristic value of the feature according to described image sample and the feature
Enclose and optimal value is calculated the step of the machine of feature of described image sample scores and included:
Judge the characteristic value of feature of described image sample whether in [Vai, Vbi] interval outer;
When described image sample feature characteristic value in [Vai, Vbi] interval outer, the then machine of the feature of described image sample
Device scoring is equal to 1;
When described image sample feature characteristic value in [Vai, Vbi] it is interval in, then the machine of the feature of described image sample
Device scoring calculation formula beWherein, ViRepresent the feature i of described image sample machine scoring;FiTable
Show the feature i of described image sample characteristic value.
Alternatively, the feature of the composition aesthetic factors includes:The ratio of marking area and whole image size and aobvious
Regional center o'clock is write to four golden section points apart from sum;
The feature of the brightness aesthetic factors includes:Image each pixel V channel brightness value sums in HSV space are returned
Average dark channel value, marking area and background area brightness value in one change value, image artwork area after standardization
The normalization characteristic value of ratio, the normalization characteristic value of the difference of marking area and background area brightness value, marking area with
Normalization characteristic value, marking area and the normalization characteristic of the difference of whole image brightness values of the ratio of whole image brightness values
Value, the average brightness value of marking area, the luminance difference and the ratio of marking area area of background area and marking area;
The feature of the color aesthetic factors includes:The overall situation of image average tone value, the global average staturation of image
Value, the bar number based on hue histogram, pixel number in the most high frequency vertical bar based on hue histogram, based on hue histogram
Maximum Hue difference, the bar number based on saturation histogram, pixel in the most high frequency vertical bar based on saturation histogram
Number, the average tone value of the maximum saturation difference based on saturation histogram, marking area, marking area average staturation value,
The saturation degree of the Hue difference of background area and marking area and the ratio of marking area area and background area and marking area
Difference and the ratio of marking area area;
The feature that the depth of field aesthetic factors include includes:The depth of field of H passages, the depth of field of channel S, the depth of field of V passages and
The ratio of the Wavelet transformation of marking area and the Wavelet transformation of background area.
Alternatively, it is described that the qualitative of aesthetic factors of described image sample is judged according to each mark person, calculate each mark
The step of person is to the artificial scorings of the aesthetic factors of described image sample includes:
Obtain mark person respectively to judge the qualitative of aesthetic factors, the aesthetic factors include composition aesthetic factors, brightness
Aesthetic factors, color aesthetic factors and depth of field aesthetic factors, it is described it is qualitative judge including it is good, in and it is poor;
It is wherein, qualitative that the artificial scoring of the aesthetic factors is 1 when judging preferably, it is qualitative judge for it is middle when, the aesthetic factors
Artificial scoring be 5, qualitative when judging for difference, the artificial scoring of the aesthetic factors is 10.
Alternatively, according to described image sample aesthetic factors machine scoring, the scoring weight of each mark person and each mark
Artificial scoring of the note person to the aesthetic factors of described image sample, calculates the step of the overall score of the aesthetic factors of described image sample
Suddenly include:
The overall score of the aesthetic factors of described image sampleWherein, its
In, SjRepresent the aesthetic factors j of described image sample overall score, gjRepresent that the aesthetic factors j of described image sample machine is commented
Point, N indicates the N mark manually scored persons, βlThe scoring weight of the l mark manually scored persons is represented,
RljRepresent artificial scoring of the l mark manually scored the persons to the aesthetic factors j of described image sample.
Alternatively, the judged result inputted according to multiple mark persons, calculates the overall aesthetic quality of described image sample
The step of scoring, includes:
Obtain whether the liking image of multiple mark persons input, whether clear, three problems whether having a distinct theme
Judged result;
Using the identical judged result more than half of each problem as the final judged result of the problem, obtain three and ask
The final judged result of topic;
It regard the total score of the final judged result of three problems as the overall aesthetic quality of described image sample
Scoring, wherein, final judged result when final judged result is no, is scored at 10 when being, to be scored at 1.
Alternatively, it is that high/low composition is beautiful by image labeling according to the overall score of the composition aesthetic factors of described image sample
Feel quality image, according to the overall score of the brightness aesthetic factors of described image sample by image labeling be high/low brightness aesthetic feeling product
Matter image, according to the overall score of the color aesthetic factors of described image sample by image labeling be high/low color aesthetic feeling quality figure
Picture, according to the overall score of the depth of field aesthetic factors of described image sample by image labeling be high/low depth of field aesthetic feeling quality image and
According to the scoring of the overall aesthetic quality of described image sample by image labeling be the high/low depth of field aesthetic feeling quality image the step of wrap
Include:
The median of the overall score of the composition aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5,
Then first threshold is equal to 5, if the median is less than 5, regard the median as first threshold;
The image pattern that the overall score of composition aesthetic factors is more than the first threshold is labeled as low composition aesthetic feeling quality
Image, high composition aesthetic feeling product are labeled as by the image pattern that the overall score of composition aesthetic factors is less than or equal to the first threshold
Matter image;
The median of the overall score of the brightness aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5,
Then Second Threshold is equal to 5, if the median is less than 5, regard the median as Second Threshold;
The image pattern that the overall score of brightness aesthetic factors is more than the Second Threshold is labeled as low-light level aesthetic feeling quality
Image, high brightness aesthetic feeling product are labeled as by the image pattern that the overall score of brightness aesthetic factors is less than or equal to the Second Threshold
Matter image;
The median of the overall score of the color aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5,
Then the 3rd threshold value is equal to 5, if the median is less than 5, regard the median as the 3rd threshold value;
The image pattern that the overall score of color aesthetic factors is more than the 3rd threshold value is labeled as low color aesthetic feeling quality
Image, high color aesthetic feeling product are labeled as by the image pattern that the overall score of color aesthetic factors is less than or equal to the 3rd threshold value
Matter image;
The median of the overall score of the depth of field aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5,
Then the 4th threshold value is equal to 5, if the median is less than 5, regard the median as the 4th threshold value;
The image pattern that the overall score of depth of field aesthetic factors is more than the 4th threshold value is labeled as low depth of field aesthetic feeling quality
Image, high depth of field aesthetic feeling product are labeled as by the image pattern that the overall score of depth of field aesthetic factors is less than or equal to the 4th threshold value
Matter image;
The median of the scoring of the overall aesthetic quality of multiple images sample is obtained, if the median is more than or equal to 5,
5th threshold value is equal to 5, if the median is less than 5, regard the median as the 5th threshold value;
The scoring of overall aesthetic quality is labeled as low overall aesthetic quality figure more than the image pattern of the 5th threshold value
Picture, high overall aesthetic quality figure is labeled as by the scoring of overall aesthetic quality less than or equal to the image pattern of the 5th threshold value
Picture.
The database building method that the present invention is provided, according to the scoring of the machine of aesthetic factors, the scoring weight of each mark person
And artificial scoring of each mark person to aesthetic factors, the overall score of the aesthetic factors of described image sample is calculated, according to multiple
The judged result of mark person input, calculates the scoring of the overall aesthetic quality of described image sample, and according to first threshold, second
Image pattern is labeled as high/low composition aesthetic feeling quality image by threshold value, the 3rd threshold value, the 4th threshold value and the 5th threshold value, high/low bright
Spend aesthetic feeling quality image, high/low color aesthetic feeling quality image, high/low depth of field aesthetic feeling quality image and high/low overall aesthetic quality
Image, realizes the foundation of database.The database that the database building method provided using the present invention is set up can be used for image
The careful research of aesthetic feeling quality.
Brief description of the drawings
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be attached to what is used required in embodiment
Figure is briefly described.It should be appreciated that the following drawings illustrate only certain embodiments of the present invention, therefore it is not construed as pair
The restriction of scope, for those of ordinary skill in the art, on the premise of not paying creative work, can also be according to this
A little accompanying drawings obtain other related accompanying drawings.
The application scenarios for the database that Fig. 1 sets up for a kind of database building method that present pre-ferred embodiments are provided show
It is intended to.
A kind of flow chart for database building method that Fig. 2 provides for present pre-ferred embodiments.
Fig. 3 is the schematic diagram of sub-step that step S120 shown in Fig. 2 includes in an embodiment.
Fig. 4 is the schematic diagram of sub-step that step S120 shown in Fig. 2 includes in another embodiment.
Fig. 5 is the schematic diagram of sub-step that sub-step S121 shown in Fig. 3 and Fig. 4 includes in an embodiment.
Fig. 6 is the schematic diagram of sub-step that step S160 shown in Fig. 2 includes in an embodiment.
Fig. 7 is the schematic diagram of sub-step that step S170 shown in Fig. 2 includes in an embodiment.
Icon:100- electronic equipments;110- memories;120- processors;130- mixed-media network modules mixed-medias.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation is described.Obviously, described embodiment is only a part of embodiment of the present invention, rather than whole embodiments.It is logical
The component for the embodiment of the present invention being often described and illustrated herein in the accompanying drawings can be arranged and designed with a variety of configurations.
Therefore, the detailed description of embodiments of the invention below to providing in the accompanying drawings is not intended to limit claimed
The scope of the present invention, but be merely representative of the present invention selected embodiment.Based on embodiments of the invention, people in the art
The every other embodiment that member is obtained on the premise of creative work is not made, belongs to the scope of protection of the invention.
It should be noted that:Similar label and letter represents similar terms in following accompanying drawing, therefore, once a certain Xiang Yi
It is defined in individual accompanying drawing, then it further need not be defined and explained in subsequent accompanying drawing.In description of the invention
In, term " first ", " second ", " the 3rd ", " the 4th " etc. are only used for distinguishing description, and it is not intended that being or implying relative
Importance.
Referring to Fig. 1, being the applied field of the database for the database building method foundation that present pre-ferred embodiments are provided
Scape schematic diagram.Database in the embodiment of the present invention can be applied in the electronic equipment 100 of evaluation image aesthetic feeling quality.The electricity
Sub- equipment 100 can possess the equipment of data-handling capacity for server, computer etc..As shown in figure 1, electronic equipment 100 is wrapped
Include:Memory 110, processor 120 and mixed-media network modules mixed-media 130.
The memory 110, processor 120 and mixed-media network modules mixed-media 130 are directly or indirectly electrically connected with each other, with
Realize the transmission or interaction of data.For example, these elements each other can be real by one or more communication bus or signal wire
Now it is electrically connected with.Memory 110, which includes at least one, to be stored in the storage in the form of software or firmware (firmware)
Software function module in device 110, the processor 120 is stored in software program and mould in memory 110 by operation
Block, so as to perform various function application and data processing, that is, implements the database building method in the embodiment of the present invention.
Wherein, the memory 110 may be, but not limited to, random access memory (Random Access
Memory, RAM), read-only storage (Read Only Memory, ROM), programmable read only memory (Programmable
Read-Only Memory, PROM), erasable read-only memory (Erasable Programmable Read-Only
Memory, EPROM), electricallyerasable ROM (EEROM) (Electric Erasable Programmable Read-Only
Memory, EEPROM) etc..Wherein, memory 110 be used for storage program, the processor 120 after execute instruction is received,
Perform described program.
The processor 120 is probably a kind of IC chip, the disposal ability with signal.Above-mentioned processor
120 can be general processor, including central processing unit (Central Processing Unit, CPU), network processing unit
(Network Processor, NP) etc..Can also be digital signal processor (DSP)), application specific integrated circuit (ASIC), scene
Programmable gate array (FPGA) or other PLDs, discrete gate or transistor logic, discrete hardware group
Part.It can realize or perform each method, step and the logic diagram disclosed in the embodiment of the present invention.General processor can be
Microprocessor or the processor 120 can also be any conventional processors etc..
Mixed-media network modules mixed-media 130 is used for the communication connection set up by network between electronic equipment 100 and external communications terminals, real
The transmitting-receiving operation of existing network signal and data.Above-mentioned network signal may include wireless signal or wire signal.
It is appreciated that the structure shown in Fig. 1 be only signal, electronic equipment 100 may also include it is more more than shown in Fig. 1 or
Less component, or with the configuration different from shown in Fig. 1.Each component shown in Fig. 1 can using hardware, software or its
Combination is realized.
Referring to Fig. 2, being a kind of flow chart for database building method that present pre-ferred embodiments are provided.The database
The database that method for building up is set up can be used for evaluation image aesthetic feeling quality.Method defined in the relevant flow of methods described is walked
Suddenly it can be realized by the processor 120.The idiographic flow shown in Fig. 2 will be described in detail below.
Step S110, obtains the multiple image of multi-class, different qualities, to be used as image pattern.
The multiple image of multi-class, different qualities can be collected by network, image pattern is used as.
Step S120, calculates the machine scoring of the aesthetic factors of described image sample.
Wherein, the aesthetic factors include composition aesthetic factors, brightness aesthetic factors, color aesthetic factors and depth of field aesthetic feeling
Factor.
Referring to Fig. 3, alternatively, step S120 includes sub-step S121, sub-step S123 and sub-step S125.Or,
Referring to Fig. 4, step S120 includes sub-step S121, sub-step S123 and sub-step S127.
Sub-step S121, calculates the machine scoring of the feature of the aesthetic factors of described image sample.
Alternatively, the feature of the composition aesthetic factors includes:The ratio of marking area and whole image size and aobvious
Regional center o'clock is write to four golden section points apart from sum.
Describe for convenience, the ratio of marking area and whole image size is designated as feature F1, by marking area center
O'clock it is designated as feature F apart from sum to four golden section points2。
So, the feature of the composition aesthetic factors includes:Feature F1With feature F2。
Alternatively, the feature of the brightness aesthetic factors includes:Image each pixel V channel brightness values in HSV space
Average dark channel value, marking area and background area on the normalized value of sum, image artwork area after standardization
It is the normalization characteristic value of the difference of the normalization characteristic value of the ratio of domain brightness value, marking area and background area brightness value, aobvious
The normalization characteristic value of region and the ratio of whole image brightness values, marking area is write with the difference of whole image brightness values to return
One changes characteristic value, the average brightness value of marking area, the luminance difference of background area and marking area and marking area area
Ratio.
By image, the normalized value of each pixel V channel brightness value sums is designated as feature F in HSV space3, image is former
Average dark channel value on the area of pictural surface after standardization is designated as feature F4, by marking area and background area brightness value
The normalization characteristic value of ratio is designated as feature F5, the normalization characteristic value of marking area and the difference of background area brightness value is remembered
It is characterized F6, the normalization characteristic value of marking area and the ratio of whole image brightness values is designated as feature F7, by marking area with
The normalization characteristic value of the difference of whole image brightness values is designated as feature F8, the average brightness value of marking area is designated as feature F9,
The ratio of the luminance difference of background area and marking area and marking area area is designated as feature F10。
So, the feature of the brightness aesthetic factors includes:Feature F3, feature F4, feature F5, feature F6, feature F7, feature
F8, feature F9With feature F10。
The feature of the color aesthetic factors includes:The overall situation of image average tone value, the global average staturation of image
Value, the bar number based on hue histogram, pixel number in the most high frequency vertical bar based on hue histogram, based on hue histogram
Maximum Hue difference, the bar number based on saturation histogram, pixel in the most high frequency vertical bar based on saturation histogram
Number, the average tone value of the maximum saturation difference based on saturation histogram, marking area, marking area average staturation value,
The saturation degree of the Hue difference of background area and marking area and the ratio of marking area area and background area and marking area
Difference and the ratio of marking area area.
The average tone value of the overall situation of image is designated as feature F11, the global average staturation value of image is designated as feature F12,
Bar number scale based on hue histogram is characterized F13, pixel number in the most high frequency vertical bar based on hue histogram is designated as
Feature F14, the maximum Hue difference based on hue histogram is designated as feature F15, it is by the bar number scale based on saturation histogram
Feature F16, pixel number in the most high frequency vertical bar based on saturation histogram is designated as feature F17, saturation degree Nogata will be based on
The maximum saturation difference of figure is designated as feature F18, the marking area tone value that is averaged is designated as feature F19, marking area is averagely satisfied
Feature F is designated as with angle value20, the ratio of the Hue difference of background area and marking area and marking area area is designated as feature
F21And the ratio of the saturation degree difference of background area and marking area and marking area area is designated as feature F22。
So, the feature of the color aesthetic factors includes:Feature F11, feature F12, feature F13, feature F14, feature F15、
Feature F16, feature F17, feature F18, feature F19, feature F20, feature F21With feature F22。
The feature that the depth of field aesthetic factors include includes:The depth of field of H passages, the depth of field of channel S, the depth of field of V passages and
The ratio of the Wavelet transformation of marking area and the Wavelet transformation of background area.
The depth of field of H passages is designated as feature F23, the depth of field of channel S is designated as feature F24, the depth of field of V passages is designated as spy
Levy F25And the ratio of the Wavelet transformation of marking area and the Wavelet transformation of background area is designated as feature F26。
So, the feature of the color aesthetic factors includes:Feature F23, feature F24, feature F25With feature F26。
Referring to Fig. 5, alternatively, sub-step S121 includes sub-step S1211, sub-step S1213 and sub-step S1215.
Sub-step S1211, calculates the characteristic value of the feature of the aesthetic factors of described image sample.
The feature of the aesthetic factors of described image sample can include feature F1To feature F26.Then calculate described image sample
Aesthetic factors feature characteristic value, for calculate feature F1To feature F26Value.
Calculating feature F1To feature F26Value when, it is necessary to extract the marking area of image.Alternatively, in the present embodiment
In, the marking area of image is extracted using the marking area extracting method based on GBVS.This method be it is a kind of it is brand-new from
The upward vision significance system in bottom, its core is following 2 points:First, the central feature passage of image is built relatively
The dynamic mapping answered;Second, the dynamic mapping obtained in the first step is entered using highly significant and the method for mapping mixing
Row standardization.
Feature F1Calculation formula be:
In formula, SARepresent the area of the marking area of image;W represents the width of whole image;H represents the height of whole image.
Feature F2Calculation formula be:
In formula, (xsc, ysc) represent marking area central point coordinate;(xi, yi) represent image four golden section points
In a point coordinate.
Feature F3Calculation formula be:
In formula, VIRepresent image I total luminance value, i.e. image I each pixel V channel brightness value sums in HSV space.
Feature F4Calculation formula be:
In formula, Idark(i) image I dark is represented;Ω (i) represents a window centered on pixel i;Ic(i) table
Diagram is as I each passage.
Feature F5Calculation formula be:
In formula, VIsRepresent the total luminance value of image I marking area;VIbRepresent the total brightness of image I background area
Value, wherein, background area is the non-significant region in whole image.
Feature F6Calculation formula be:
Feature F7Calculation formula be:
Feature F8Calculation formula be:
Feature F9Calculation formula be:
Feature F10Calculation formula be:
Feature F11Calculation formula be:
In formula, IH(x, y) represents the tone value of each pixel in image I.
Feature F12Calculation formula be:
In formula, IH(x, y) represents the intensity value of each pixel in image I.
Feature F13Calculation formula be:
In formula,Expression group is away from 20 hue histogram;Q represents the maximum of hue histogram;C joins for acquiescence
Number, its value is 0.1.
Carrying out calculating feature F13When, it is first depending on condition IS> 0.2 and 0.95>IL>0.15 pair of image pixel is sieved
Choosing, so that avoid saturation degree too low, and the influence that the too high and too low pixel of brightness is counted to color characteristic, then to passing through
The tone value of pixel carries out the statistics with histogram at intervals of 20 after screening, obtains hue histogram.
Feature F14Calculation formula be:
In formula,Represent the maximum of hue histogram, i.e., tone frequency of occurrences highest picture in image
Plain number.
Feature F15Calculation formula be:
In formula,Represent the maximum hue difference in hue histogram.
Similarly, it can also be drawn based on the color characteristic that saturation histogram is carried by the above method and feature F13, feature F14
With feature F15Similar feature F16, feature F17With feature F18Calculation formula.
Feature F16Calculation formula be:
In formula,Expression group is away from 20 saturation histogram;The maximum of Q ' expression saturation histograms;C ' is acquiescence
Parameter, its value is 0.1.
Feature F17Calculation formula be:
In formula,Represent the maximum of saturation histogram, i.e., saturation degree frequency of occurrences highest in image
Number of pixels.
Feature F18Calculation formula be:
In formula,Represent that the maximum saturation in hue histogram is poor.
Feature F19Calculation formula be:
In formula, hIsRepresent total tone value of image I marking area.
Feature F20Calculation formula be:
In formula, sIsRepresent total intensity value of image I marking area.
Feature F21Calculation formula be:
In formula, hIbRepresent total tone value of image I background area.
Feature F22Calculation formula be:
In formula, sIbRepresent total intensity value of image I background area.
Feature F23Calculation formula be:
In formula, ω3H(x, y) represents the three-level wavelet coefficient sum of marking area high frequency on H passages;The three-level wavelet coefficient of marking area high frequency on H passages is represented respectively, and wavelet transform procedure is as schemed
Shown in 3;SalientArea represents marking area;AllArea represents whole image.
Similarly, the depth of field feature for obtaining S, V can be calculated on S, V passage, feature F is released24With feature F25Calculating it is public
Formula.
Feature F24Calculation formula be:
In formula, ω3S(x, y) represents the three-level wavelet coefficient sum of marking area high frequency in channel S;The three-level wavelet coefficient of marking area high frequency in channel S is represented respectively;SalientArea is represented
Marking area;AllArea represents whole image.
Feature F25Calculation formula be:
In formula, ω3V(x, y) represents the three-level wavelet coefficient sum of marking area high frequency on V passages;The three-level wavelet coefficient of marking area high frequency on V passages is represented respectively;SalientArea is represented
Marking area;AllArea represents whole image.
Feature F26Calculation formula be:
In formula, molecule represents sum of the three-level wavelet transformation in tri- passages of H, S, V of marking area;Denominator represents background area
Sum of the three-level wavelet transformation in domain in tri- passages of H, S, V.
Feature F can just be calculated using above-mentioned formula1To feature F26Characteristic value.
Sub-step S1213, according to the characteristic value of the feature of multiple high aesthetic feeling quality images, calculates the high aesthetic feeling of the feature
Quality span and optimal value.
Alternatively, sub-step S1213 includes sub-step S12131, sub-step S12133 and sub-step S12135.
Sub-step S12131, according to the characteristic value of each feature of multiple high aesthetic feeling quality images, draws each feature straight respectively
Fang Tu, wherein, the width of the Nogata post of each feature histogram uses Scott derivation formulas, each feature histogram
Ordinate represents the picture number of the high aesthetic feeling quality image corresponding to the Nogata post, and abscissa is represented corresponding to the Nogata post
The characteristic value of feature.
Substantial amounts of multi-class, high aesthetic feeling quality image can be collected by network, high aesthetic feeling quality image sample is used as.
According to above-mentioned formula, the feature F of multiple high aesthetic feeling quality images is calculated1To feature F26Characteristic value, draw feature F1To spy
Levy F26Histogram, obtains 26 feature histograms.The width of the Nogata post of each feature histogram uses Scott derivation formulas,
The histogrammic ordinate of this feature represents the picture number of the high aesthetic feeling quality image corresponding to the Nogata post, and abscissa represents this
The characteristic value of feature corresponding to Nogata post.
For example, the high aesthetic feeling quality image has n, the feature F of the n high aesthetic feeling quality images1Characteristic value
Average is μ.Draw feature F1Histogram, makes this feature F1Histogrammic width hn, then hnCalculation formula be:
Histogrammic width h is tried to achieve using above-mentioned formulanAfterwards, with feature F1Characteristic value as abscissa, sat so that correspondence is horizontal
The picture number of the high aesthetic feeling quality image of target draws feature F as ordinate1Histogram.
Sub-step S12133, calculates the extreme difference of the ordinate of each feature histogram, obtains in each feature histogram and indulges
Coordinate is more than or equal to the abscissa scope of the Nogata post of 0.5 times of extreme difference, is used as the high aesthetic feeling quality value model of each feature
Enclose, be designated as [Vai, Vbi]。
Wherein, i can be equal to 1,2,3 ..., and 26.Feature F1High aesthetic feeling quality span be [Va1, Vb1], feature F2
High aesthetic feeling quality span be [Va2, Vb2].Similarly, feature F is obtained3To feature F26High aesthetic feeling quality span.
With feature F1Exemplified by histogram, this feature F1Histogrammic ordinate maximum is ymax, ordinate minimum value is
ymin, then this feature F1The extreme difference of histogrammic ordinate is (ymax-ymin).Selected characteristic F1Ordinate is more than or waited in histogram
In 0.5 (ymax-ymin) Nogata post where abscissa scope, be used as feature F1High aesthetic feeling quality span, be designated as
[Va1, Vb1]。
Sub-step S12135, obtains the Nogata that ordinate in each feature histogram is more than or equal to 0.9 times of maximum
The intermediate value of the abscissa zone of post is designated as V as the optimal value of each featureopti。
Similarly, feature F1Optimal value be Vopt1, feature F2Optimal value be Vopt2... feature F26Optimal value is Vopt26。
With feature F1Exemplified by histogram, this feature F1Histogrammic ordinate is more than or equal to 0.9ymaxNogata post horizontal stroke
Coordinate is interval, regard the intermediate value of the abscissa zone as this feature F1Optimal value, be designated as Vopt1。
Sub-step S1215, according to the characteristic value of the feature of described image sample and the high aesthetic feeling quality value of the feature
Scope and optimal value calculate the machine scoring of the feature of described image sample.
Alternatively, sub-step S1215 includes sub-step S12151 and sub-step S12153.Or, sub-step S1215 includes
Sub-step S12151 and sub-step S12155.
Whether sub-step S12151, judge the characteristic value of feature of described image sample in [Vai, Vbi] interval outer.
The feature of described image sample includes feature F1To feature F26.Judging characteristic F successively1Characteristic value whether in [Va1,
Vb1] interval outer, feature F2Characteristic value whether in [Va2, Vb2] interval outer ..., feature F26Characteristic value whether in [Va26, Vb26]
It is interval outer.
As the characteristic value F of the feature of described image sampleiIn [Vai, Vbi] interval outer, perform sub-step S12153.When described
The characteristic value of the feature of image pattern is in [Vai, Vbi] interval interior, perform sub-step S12155.
Sub-step S12153, the machine scoring of the feature of described image sample is equal to 1.
For example, as feature F1Characteristic value in [Va1, Vb1] it is interval outer when, the feature F of the image pattern1Machine scoring etc.
In 1.As feature F2Characteristic value in [Va2, Vb2] it is interval outer when, the feature F of the image pattern2Machine scoring be equal to 1.
Sub-step S12155, the feature of described image sample machine scoring calculation formula beIts
In, ViRepresent the feature i of described image sample machine scoring;FiRepresent the feature i of described image sample characteristic value.
For example, feature F1Characteristic value in [Va1, Vb1] it is interval in when, then the feature F of the image pattern1Machine scoringFeature F2Characteristic value in [Va2, Vb2] it is interval in when, then the feature F of the image pattern2Machine scoring
Similarly, the feature F of image pattern can be calculated3To feature F26Machine scoring.
Sub-step S123, calculates the average value of the machine scoring of the feature of the aesthetic factors of described image sample.
Because aesthetic factors include:Composition aesthetic factors, brightness aesthetic factors, color aesthetic factors and depth of field aesthetic feeling because
Element.So, sub-step S123 is the average value of the machine scoring of the feature of calculating composition aesthetic factors respectively, brightness aesthetic factors
Feature machine scoring average value, the feature of color aesthetic factors machine scoring average value, depth of field aesthetic factors
The average value of the machine scoring of feature
Because the feature of composition aesthetic factors includes feature F1With feature F2, the features of brightness aesthetic factors includes feature F3
To feature F10, the features of color aesthetic factors includes feature F11To feature F22, the features of depth of field aesthetic factors includes feature F22Extremely
Feature F26.So, the average value of the machine scoring of the feature of composition aesthetic factors is characterized F1With feature F2The machine of this 2 features
The average value of device scoring.The average value of the machine scoring of the feature of brightness aesthetic factors is characterized F3To feature F10This 8 features
Machine scoring average value.The average value of the machine scoring of the feature of color aesthetic factors is characterized F11To feature F22This 12
The average value of the machine scoring of individual feature.The average value of the machine scoring of the feature of depth of field aesthetic factors is characterized F23To feature
F26The average value of the machine scoring of this 4 features.
When the average value of the machine scoring of the feature of the aesthetic factors of described image sample is less than 0.1, sub-step is performed
S125, when the average value of the machine scoring of the feature of the aesthetic factors of described image sample is more than or equal to 0.1, performs sub-step
Rapid S127.
Sub-step S125, the machine scoring of the aesthetic factors of described image sample is equal to 1.
For example, as feature F1With feature F2When the average value of the machine scoring of this 2 features is less than 0.1, described image sample
Composition aesthetic factors machine scoring be equal to 1.As feature F3To feature F10The average value of the machine scoring of this 8 features is less than
When 0.1, the machine scoring of the brightness aesthetic factors of described image sample is equal to 1.
Sub-step S127, the machine scoring of the aesthetic factors of described image sample is equal to 10 times of average values.
For example, as feature F1With feature F2When the average value of the machine scoring of this 2 features is more than or equal to 0.1, the figure
The machine scoring of the composition aesthetic factors of decent is equal to 10 times of feature F1With feature F2The machine scoring of this 2 features
Average value.As feature F3To feature F10When the average value of the machine scoring of this 8 features is more than or equal to 0.1, described image sample
The machine scoring of this brightness aesthetic factors is equal to 10 times of feature F3To feature F10What the machine of this 8 features scored is averaged
Value.
Similarly, using sub-step S125 or sub-step S127 can calculate described image sample color aesthetic factors and
The machine scoring of depth of field aesthetic factors.
Step S130, survey is carried out to mark person, and the problem of being inputted according to each mark person answer calculates mark person
Scoring weight.
It is the proportion in overall score that determines manually to score, it is necessary to the fine arts, the photography work(of the mark person to participating in scoring
Bottom and estheticism carry out survey.Investigation problem can have 10, and the score value of each problem is 0.1 point, according to each mark
The problem of person inputs answer, calculating the formula of the scoring weight of mark person can be:
Wherein, SHlRepresent that the l mark manually scored persons answer the questionnaire score of investigation problem, βlRepresent l
The scoring weight for the mark person that name is manually scored.
The investigation problem can include:1. whether movies-making was learnt2. whether art painting was learnt3. whether
Learnt appreciation of arts4. two best images of composition are selected from following six width figure5. selected from following six width figure bright
Two best images of degree6. two best images of color are selected from following six width figure7. selected from following six width figure
Two best images of the depth of field8. which two image you are more likely in following six width figures9. which two figures in six width figures below
Image sharpness is more preferable10. which two image is the animal painting having a distinct theme in following six width figures
Step S140, judges according to each mark person to the qualitative of aesthetic factors of described image sample, calculates each mark person
Artificial scoring to the aesthetic factors of described image sample.
Artificial scoring to the aesthetic factors of described image sample is included to composition aesthetic factors, brightness aesthetic factors, face
The artificial scoring of color aesthetic factors and depth of field aesthetic factors.When manually being scored, mark is considered in the embodiment of the present invention
Person can not possibly make very careful scoring to image, therefore only need mark person enters to the composition of image, brightness, color and the depth of field
Row " good ", " in ", the qualitative evaluation of " poor ".Wherein the artificial scoring of " good " corresponding aesthetic factors can be 1, " in " corresponding
The artificial scoring of aesthetic factors can be 5, and the artificial scoring of " poor " corresponding aesthetic factors can be 10.
Often occur the skimble-scamble situation of opinion, the i.e. same aesthetic factors to same image pattern during artificial scoring to carry out
During scoring, some mark persons are judged as " good ", and other marks person is judged as " poor ".In the embodiment of the present invention, by quantity
The evaluation result of a side on the low side regards as extreme event.In order to ensure that the scoring that each mark person provides can be utilized,
Simultaneously avoid in the influence that extreme event is brought, the embodiment of the present invention, by extreme event be changed into intermediate form " in ".Using
Such a design, not only can carry out making full use of but also eliminating the influence that extreme event is brought, keep simultaneously to the judge of mark person
The generality of event.
For example, there is 10 mark persons to participate in artificial scoring, a certain aesthetic factors that 7 mark persons judge the image pattern are
" good ", 2 mark persons judge a certain aesthetic factors of the image pattern for " in ", 1 mark person judges certain of the image pattern
One aesthetic factors are " poor ", then this " poor " are regarded as into extreme event, " poor " is changed into " in ".
Step S150, according to the machine of the aesthetic factors of described image sample scoring, the scoring weight of each mark person and
Artificial scoring of each mark person to the aesthetic factors of described image sample, calculates the general comment of the aesthetic factors of described image sample
Point.
Calculating the overall score of the aesthetic factors of described image sample includes calculating the total of the composition aesthetic factors of image pattern
The overall score of scoring, the overall score of brightness aesthetic factors, the overall score of color aesthetic factors and depth of field aesthetic factors..
The overall score of the aesthetic factors of described image sample
Wherein, SjRepresent the aesthetic factors j of described image sample overall score, gjRepresent described image sample aesthetic feeling because
Plain j machine scoring.As j=1, S1Represent the overall score of the composition aesthetic factors of described image sample, g1Represent described image
The machine scoring of the composition aesthetic factors of sample.As j=2, S2Represent the general comment of the brightness aesthetic factors of described image sample
Point, g2Represent the machine scoring of the brightness aesthetic factors of described image sample.As j=3, S3Represent the face of described image sample
The overall score of color aesthetic factors, g3Represent the machine scoring of the color aesthetic factors of described image sample.As j=4, S4Represent
The overall score of the depth of field aesthetic factors of described image sample, g4Represent that the machine of the depth of field aesthetic factors of described image sample is commented
Point.
N indicates the N mark manually scored persons, βlRepresent the scoring of the l mark manually scored persons
Weight.RljRepresent artificial scoring of the l mark manually scored the persons to the aesthetic factors j of described image sample.Work as j=
When 1, RliRepresent artificial scoring of the l mark manually scored the persons to the composition aesthetic factors of described image sample.Work as j
When=2, Rl2Represent artificial scoring of the l mark manually scored the persons to the brightness aesthetic factors of described image sample.
As j=3, Rl3Represent that the l mark manually scored persons are commented the artificial of color aesthetic factors of described image sample
Point.As j=4, Rl4Represent the l mark manually scored persons to the artificial of the depth of field aesthetic factors of described image sample
Scoring.
Step S160, the judged result inputted according to multiple mark persons calculates the overall aesthetic quality of described image sample
Scoring.
Referring to Fig. 6, alternatively, step S160 includes sub-step S161, sub-step S163 and sub-step S165.
Sub-step S161, obtain whether the liking image of multiple mark persons input, it is whether clear, whether have a distinct theme
The judged result of three problems.
Sub-step S163, using the identical judged result more than half of each problem as the problem final judgement knot
Really, the final judged result of three problems is obtained.
It is if more than half mark persons obtain identical judged result, this is identical for the same problem of same image pattern
Judged result is used as final judged result.For example, totally five mark persons participate in scoring, they are for above three question answering point
It is not " being whether ", " being whether ", " whether no ", " being ", " being ", then according to the principle that the minority is subordinate to the majority, most
Whole judged result is " being whether ".
Sub-step S165, using the total score of the final judged result of three problems as described image sample entirety
The scoring of aesthetic feeling quality, wherein, final judged result when final judged result is no, is scored at 10 when being, to be scored at 1.
For example, final judged result is " being ", the scoring for assert the overall aesthetic quality of the image pattern is 1+1+1
=3 points.Final judged result is " being whether ", and the scoring for assert the overall aesthetic quality of the image pattern is 1+1+10=12
Point.Whether final judged result is " no ", and the scoring for assert the overall aesthetic quality of the image pattern is 1+10+10=21 points.
Final judged result is " no no ", and the scoring for assert the overall aesthetic quality of the image pattern is 10+10+10=30 points.It is logical
Cross above three problem and the overall aesthetic quality of image is divided into 3 points, 12 points, 21 points and 30 points four grades.Fraction is lower, when
The overall aesthetic quality of preceding image pattern is higher.
Step S170, according to the overall score of the composition aesthetic factors of described image sample by image labeling be high/low composition
Aesthetic feeling quality image, according to the overall score of the brightness aesthetic factors of described image sample by image labeling be high/low brightness aesthetic feeling
Quality image, according to the overall score of the color aesthetic factors of described image sample by image labeling be high/low color aesthetic feeling quality
Image, according to the overall score of the depth of field aesthetic factors of described image sample by image labeling be high/low depth of field aesthetic feeling quality image
And according to the scoring of the overall aesthetic quality of described image sample by image labeling be high/low depth of field aesthetic feeling quality image, to build
Vertical database.
Referring to Fig. 7, alternatively, step S170 includes sub-step S171, sub-step S172, sub-step S173, sub-step
S174, sub-step S175, sub-step S176, sub-step S177, sub-step S178, sub-step S179 and sub-step S180.
Sub-step S171, obtains the median of the overall score of the composition aesthetic factors of multiple images sample, if the median
More than or equal to 5, then first threshold is equal to 5, if the median is less than 5, regard the median as first threshold.
Sub-step S172, low structure is labeled as by the image pattern that the overall score of composition aesthetic factors is more than the first threshold
Figure aesthetic feeling quality image, height is labeled as by the image pattern that the overall score of composition aesthetic factors is less than or equal to the first threshold
Composition aesthetic feeling quality image.
Sub-step S173, obtains the median of the overall score of the brightness aesthetic factors of multiple images sample, if the median
More than or equal to 5, then Second Threshold is equal to 5, if the median is less than 5, regard the median as Second Threshold.
Sub-step S174, the image pattern that the overall score of brightness aesthetic factors is more than the Second Threshold is labeled as low bright
Aesthetic feeling quality image is spent, the image pattern that the overall score of brightness aesthetic factors is less than or equal to the Second Threshold is labeled as height
Brightness aesthetic feeling quality image.
Sub-step S175, obtains the median of the overall score of the color aesthetic factors of multiple images sample, if the median
More than or equal to 5, then the 3rd threshold value is equal to 5, if the median is less than 5, regard the median as the 3rd threshold value.
Sub-step S176, low face is labeled as by the image pattern that the overall score of color aesthetic factors is more than the 3rd threshold value
Color aesthetic feeling quality image, height is labeled as by the image pattern that the overall score of color aesthetic factors is less than or equal to the 3rd threshold value
Color aesthetic feeling quality image.
Sub-step S177, obtains the median of the overall score of the depth of field aesthetic factors of multiple images sample, if the median
More than or equal to 5, then the 4th threshold value is equal to 5, if the median is less than 5, regard the median as the 4th threshold value.
Sub-step S178, low scape is labeled as by the image pattern that the overall score of depth of field aesthetic factors is more than the 4th threshold value
Deep aesthetic feeling quality image, height is labeled as by the image pattern that the overall score of depth of field aesthetic factors is less than or equal to the 4th threshold value
Depth of field aesthetic feeling quality image.
Sub-step S179, obtains the median of the scoring of the overall aesthetic quality of multiple images sample, if the median is big
In or equal to 5, then the 5th threshold value is equal to 5, if the median is less than 5, regard the median as the 5th threshold value.
Sub-step S180, low entirety is labeled as by the scoring of overall aesthetic quality more than the image pattern of the 5th threshold value
Aesthetic feeling quality image, the scoring of overall aesthetic quality is labeled as less than or equal to the image pattern of the 5th threshold value high overall
Aesthetic feeling quality image.
Database building method provided in an embodiment of the present invention, proposes four aesthetic factors and four aesthetic factors bags
26 features included are used for careful research image aesthetic feeling quality, and being calculated in particular by the characteristic value of each feature of aesthetic factors should
The machine scoring of aesthetic factors, and commented according to the machine of the machine score calculation aesthetic factors corresponding with this feature of each feature
Point, and mark person is calculated to the artificial scorings of each aesthetic factors and scoring weight, machine scoring finally according to aesthetic factors,
The artificial scoring of the scoring weight of each mark person and each mark person to aesthetic factors, calculates the aesthetic factors of described image sample
Overall score, according to multiple mark persons input judged result, calculate described image sample overall aesthetic quality scoring, and
Image pattern is labeled as by high/low composition U.S. according to first threshold, Second Threshold, the 3rd threshold value, the 4th threshold value and the 5th threshold value
Feel quality image, high/low brightness aesthetic feeling quality image, high/low color aesthetic feeling quality image, high/low depth of field aesthetic feeling quality image,
With high/low overall aesthetic quality image, the foundation of database is realized.What the database building method provided using the present invention was set up
Database can be used for the careful research of image aesthetic feeling quality.
In several embodiments that the embodiment of the present invention is provided, it should be understood that disclosed method, it can also pass through
Other modes are realized.Embodiment of the method described above is only schematical, for example, flow chart and block diagram in accompanying drawing
Show the architectural frameworks in the cards of method and computer program product of multiple embodiments according to the present invention, function and
Operation.At this point, each square frame in flow chart or block diagram can represent a part for a module, program segment or code,
A part for the module, program segment or code includes one or more executable fingers for being used to realize defined logic function
Order.It should also be noted that in some implementations as replacement, the function of being marked in square frame can also be with different from accompanying drawing
Middle marked order occurs.For example, two continuous square frames can essentially be performed substantially in parallel, they sometimes can also
Perform in the opposite order, this is depending on involved function.It is also noted that each side in block diagram and/or flow chart
The combination of frame and the square frame in block diagram and/or flow chart, can be with function or action as defined in performing it is special based on hard
The system of part is realized, or can be realized with the combination of specialized hardware and computer instruction.
If the function is realized using in the form of software function module and is used as independent production marketing or in use, can be with
It is stored in a computer read/write memory medium.Understood based on such, technical scheme is substantially in other words
The part contributed to prior art or the part of the technical scheme can be embodied in the form of software product, the meter
Calculation machine software product is stored in a storage medium, including some instructions are to cause a computer equipment (can be individual
People's computer, various electronic equipments, or network equipment etc.) perform all or part of step of various embodiments of the present invention methods described
Suddenly.And foregoing storage medium includes:USB flash disk, mobile hard disk, read-only storage (ROM, Read-Only Memory), deposit at random
Access to memory (RAM, Random Access Memory), magnetic disc or CD etc. are various can be with the medium of store program codes.
It should be noted that herein, term " comprising ", "comprising" or its any other variant are intended to the bag of nonexcludability
Contain, so that process, method, article or equipment including a series of key elements are not only including those key elements, but also including
Other key elements being not expressly set out, or also include for this process, method, article or the intrinsic key element of equipment.
In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that including the key element
Process, method, article or equipment in also there is other identical element.
The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (10)
1. a kind of database building method, the database is used for image aesthetic feeling quality evaluation, it is characterised in that methods described bag
Include:
The multiple image of multi-class, different qualities is obtained, to be used as image pattern;
The machine scoring of the aesthetic factors of described image sample is calculated, wherein, the aesthetic factors include composition aesthetic factors, bright
Spend aesthetic factors, color aesthetic factors and depth of field aesthetic factors;
Survey is carried out to mark person, the problem of being inputted according to each mark person answer calculates the scoring weight of mark person;
The qualitative of aesthetic factors of described image sample is judged according to each mark person, each mark person is calculated to described image sample
Aesthetic factors artificial scoring;
According to the scoring of the machine of the aesthetic factors of described image sample, the scoring weight of each mark person and each mark person to described
The artificial scoring of the aesthetic factors of image pattern, calculates the overall score of the aesthetic factors of described image sample, including image
Overall score, the overall score of brightness aesthetic factors, the overall score of color aesthetic factors and the depth of field of the composition aesthetic factors of sample
The overall score of aesthetic factors;
The judged result inputted according to multiple mark persons, calculates the scoring of the overall aesthetic quality of described image sample;
According to the overall score of the composition aesthetic factors of described image sample by image labeling be high/low composition aesthetic feeling quality image,
According to the overall score of the brightness aesthetic factors of described image sample by image labeling be high/low brightness aesthetic feeling quality image, according to
Image labeling is high/low color aesthetic feeling quality image by the overall score of the color aesthetic factors of described image sample, according to described
The overall score of the depth of field aesthetic factors of image pattern is by image labeling for high/low depth of field aesthetic feeling quality image and according to described image
Image labeling is high/low depth of field aesthetic feeling quality image by the scoring of the overall aesthetic quality of sample, to set up database.
2. database building method according to claim 1, it is characterised in that calculate the aesthetic factors of described image sample
Machine include the step of score:
Calculate the machine scoring of the feature of the aesthetic factors of described image sample;
Calculate the average value of the machine scoring of the feature of the aesthetic factors of described image sample;
When the average value of the machine scoring of the feature of the aesthetic factors of described image sample is less than 0.1, described image sample
The machine scoring of aesthetic factors is equal to 1;
When the average value of the machine scoring of the feature of the aesthetic factors of described image sample is more than or equal to 0.1, described image
The machine scoring of the aesthetic factors of sample is equal to 10 times of average values.
3. database building method according to claim 2, it is characterised in that the aesthetic feeling of the calculating described image sample
The step of machine of the feature of factor scores includes:
Calculate the characteristic value of the feature of the aesthetic factors of described image sample;
According to the characteristic value of the feature of multiple high aesthetic feeling quality images, the high aesthetic feeling quality span and most of the feature is calculated
The figure of merit;
Calculated according to the high aesthetic feeling quality span and optimal value of the characteristic value of the feature of described image sample and the feature
The machine scoring of the feature of described image sample.
4. database building method according to claim 3, it is characterised in that according to the spy of multiple high aesthetic feeling quality images
The step of characteristic value levied, high aesthetic feeling quality span and optimal value for calculating the feature, includes:
According to the characteristic value of each feature of multiple high aesthetic feeling quality images, each feature histogram is drawn respectively, wherein, each spy
The width for levying histogrammic Nogata post uses Scott derivation formulas, and the ordinate of each feature histogram represents the Nogata post
The picture number of corresponding high aesthetic feeling quality image, abscissa represents the characteristic value of the feature corresponding to the Nogata post;
The extreme difference of the ordinate of each feature histogram is calculated, ordinate in each feature histogram is obtained and is more than or equal to 0.5
The abscissa scope of the Nogata post of the extreme difference, as the high aesthetic feeling quality span of each feature, is designated as [V againai, Vbi];
Ordinate in each feature histogram is obtained to be more than or equal in the abscissa zone of the Nogata post of 0.9 times of maximum
It is worth the optimal value as each feature, is designated as Vopti。
5. database building method according to claim 4, it is characterised in that the feature according to described image sample
Characteristic value and the high aesthetic feeling quality span and optimal value of the feature calculate the machine of feature of described image sample and comment
The step of dividing includes:
Judge the characteristic value of feature of described image sample whether in [Vai, Vbi] interval outer;
When described image sample feature characteristic value in [Vai, Vbi] interval outer, then the machine of the feature of described image sample is commented
Grade in 1;
When described image sample feature characteristic value in [Vai, Vbi] interval interior, then the machine of the feature of described image sample is commented
Point calculation formula beWherein, ViRepresent the feature i of described image sample machine scoring;FiRepresent institute
State the feature i of image pattern characteristic value.
6. database building method according to claim 2, it is characterised in that the feature bag of the composition aesthetic factors
Include:The ratio and marking area central point of marking area and whole image size are to four golden section points apart from sum;
The feature of the brightness aesthetic factors includes:The normalization of image each pixel V channel brightness value sums in HSV space
The ratio of average dark channel value, marking area and background area brightness value in value, image artwork area after standardization
The normalization characteristic value of the difference of the normalization characteristic value of value, marking area and background area brightness value, marking area with whole
The normalization characteristic value of the normalization characteristic value of the ratio of image brightness values, marking area and the difference of whole image brightness values,
The ratio of the average brightness value of marking area, the luminance difference of background area and marking area and marking area area;
The feature of the color aesthetic factors includes:The overall situation of image average tone value, the global average staturation value of image, base
Bar number in hue histogram, pixel number in the most high frequency vertical bar based on hue histogram, based on hue histogram most
Pixel number, base in big Hue difference, the bar number based on saturation histogram, the most high frequency vertical bar based on saturation histogram
Maximum saturation difference, marking area average tone value, marking area average staturation value, background area in saturation histogram
The saturation degree difference of the Hue difference of domain and marking area and the ratio of marking area area and background area and marking area with
The ratio of marking area area;
The feature that the depth of field aesthetic factors include includes:The depth of field of H passages, the depth of field of channel S, the depth of field of V passages and significantly
The ratio of the Wavelet transformation in region and the Wavelet transformation of background area.
7. database building method according to claim 1, it is characterised in that it is described according to each mark person to described image
The qualitative of the aesthetic factors of sample is judged, and calculates the step of each mark person is to the artificial scorings of the aesthetic factors of described image sample
Including:
Obtain mark person respectively to judge the qualitative of aesthetic factors, the aesthetic factors include composition aesthetic factors, brightness aesthetic feeling
Factor, color aesthetic factors and depth of field aesthetic factors, it is described it is qualitative judge including it is good, in and it is poor;
It is wherein, qualitative that the artificial scoring of the aesthetic factors is 1 when judging preferably, it is qualitative judge for it is middle when, the people of the aesthetic factors
Work scoring is 5, qualitative when judging for difference, and the artificial scoring of the aesthetic factors is 10.
8. database building method according to claim 1, it is characterised in that according to the aesthetic factors of described image sample
Machine scoring, the artificial scoring of the scoring weight of each mark person and each mark person to the aesthetic factors of described image sample,
The step of overall score for the aesthetic factors for calculating described image sample, includes:
The overall score of the aesthetic factors of described image sampleWherein, wherein, Sj
Represent the aesthetic factors j of described image sample overall score, gjRepresent the aesthetic factors j of described image sample machine scoring, N
Indicate the N mark manually scored persons, βlRepresent the scoring weight of the l mark manually scored persons, RljTable
Show artificial scoring of the l mark manually scored the persons to the aesthetic factors j of described image sample.
9. database building method according to claim 1, it is characterised in that the judgement knot inputted according to multiple mark persons
Really, the step of scoring for the overall aesthetic quality for calculating described image sample, includes:
Obtain whether the liking image of multiple mark persons input, the whether clear, judgement of three problems that whether has a distinct theme
As a result;
Using the identical judged result more than half of each problem as the final judged result of the problem, three problems are obtained
Final judged result;
Using the total score of the final judged result of three problems as the overall aesthetic quality of described image sample scoring,
Wherein, final judged result when final judged result is no, is scored at 10 when being, to be scored at 1.
10. database building method according to claim 1, it is characterised in that beautiful according to the composition of described image sample
The overall score of sense factor by image labeling be high/low composition aesthetic feeling quality image, according to the brightness aesthetic feeling of described image sample because
Image labeling is high/low brightness aesthetic feeling quality image by the overall score of element, according to the color aesthetic factors of described image sample
Image labeling is high/low color aesthetic feeling quality image by overall score, according to the general comment of the depth of field aesthetic factors of described image sample
Divide and image labeling is high/low depth of field aesthetic feeling quality image and will be schemed according to the scoring of the overall aesthetic quality of described image sample
Include as the step of being labeled as high/low depth of field aesthetic feeling quality image:
The median of the overall score of the composition aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5, the
One threshold value is equal to 5, if the median is less than 5, regard the median as first threshold;
The image pattern that the overall score of composition aesthetic factors is more than the first threshold is labeled as low composition aesthetic feeling quality image,
The image pattern that the overall score of composition aesthetic factors is less than or equal to the first threshold is labeled as high composition aesthetic feeling quality figure
Picture;
The median of the overall score of the brightness aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5, the
Two threshold values are equal to 5, if the median is less than 5, regard the median as Second Threshold;
The image pattern that the overall score of brightness aesthetic factors is more than the Second Threshold is labeled as low-light level aesthetic feeling quality image,
The image pattern that the overall score of brightness aesthetic factors is less than or equal to the Second Threshold is labeled as high brightness aesthetic feeling quality figure
Picture;
The median of the overall score of the color aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5, the
Three threshold values are equal to 5, if the median is less than 5, regard the median as the 3rd threshold value;
The image pattern that the overall score of color aesthetic factors is more than the 3rd threshold value is labeled as low color aesthetic feeling quality image,
The image pattern that the overall score of color aesthetic factors is less than or equal to the 3rd threshold value is labeled as high color aesthetic feeling quality figure
Picture;
The median of the overall score of the depth of field aesthetic factors of multiple images sample is obtained, if the median is more than or equal to 5, the
Four threshold values are equal to 5, if the median is less than 5, regard the median as the 4th threshold value;
The image pattern that the overall score of depth of field aesthetic factors is more than the 4th threshold value is labeled as low depth of field aesthetic feeling quality image,
The image pattern that the overall score of depth of field aesthetic factors is less than or equal to the 4th threshold value is labeled as high depth of field aesthetic feeling quality figure
Picture;
The median of the scoring of the overall aesthetic quality of multiple images sample is obtained, if the median is more than or equal to the 5, the 5th
Threshold value is equal to 5, if the median is less than 5, regard the median as the 5th threshold value;
The scoring of overall aesthetic quality is labeled as low overall aesthetic quality image more than the image pattern of the 5th threshold value, will
The scoring of overall aesthetic quality is labeled as high overall aesthetic quality image less than or equal to the image pattern of the 5th threshold value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710484335.9A CN107203771A (en) | 2017-06-23 | 2017-06-23 | Database building method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710484335.9A CN107203771A (en) | 2017-06-23 | 2017-06-23 | Database building method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107203771A true CN107203771A (en) | 2017-09-26 |
Family
ID=59908400
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710484335.9A Pending CN107203771A (en) | 2017-06-23 | 2017-06-23 | Database building method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107203771A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109846512A (en) * | 2019-03-20 | 2019-06-07 | 殳儆 | A kind of 12 lattice lung ultrasound image panorama formula method for reporting based on anatomical position |
CN110610479A (en) * | 2019-07-31 | 2019-12-24 | 华为技术有限公司 | Object scoring method and device |
CN110807759A (en) * | 2019-09-16 | 2020-02-18 | 幻想动力(上海)文化传播有限公司 | Method and device for evaluating photo quality, electronic equipment and readable storage medium |
-
2017
- 2017-06-23 CN CN201710484335.9A patent/CN107203771A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109846512A (en) * | 2019-03-20 | 2019-06-07 | 殳儆 | A kind of 12 lattice lung ultrasound image panorama formula method for reporting based on anatomical position |
CN110610479A (en) * | 2019-07-31 | 2019-12-24 | 华为技术有限公司 | Object scoring method and device |
CN112258450A (en) * | 2019-07-31 | 2021-01-22 | 华为技术有限公司 | Object scoring method and device |
CN110610479B (en) * | 2019-07-31 | 2024-05-03 | 花瓣云科技有限公司 | Object scoring method and device |
CN110807759A (en) * | 2019-09-16 | 2020-02-18 | 幻想动力(上海)文化传播有限公司 | Method and device for evaluating photo quality, electronic equipment and readable storage medium |
CN110807759B (en) * | 2019-09-16 | 2022-09-06 | 上海甜里智能科技有限公司 | Method and device for evaluating photo quality, electronic equipment and readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107330455A (en) | Image evaluation method | |
CN107292813B (en) | A kind of multi-pose Face generation method based on generation confrontation network | |
WO2020151489A1 (en) | Living body detection method based on facial recognition, and electronic device and storage medium | |
KR102195922B1 (en) | Internet-based facial beautification system | |
CN103745210B (en) | A kind of leucocyte classification method and device | |
CN100361450C (en) | System for blocking off erotic images and unhealthy information in internet | |
CN106469302A (en) | A kind of face skin quality detection method based on artificial neural network | |
CN107203771A (en) | Database building method | |
CN106650786A (en) | Image recognition method based on multi-column convolutional neural network fuzzy evaluation | |
CN106570474A (en) | Micro expression recognition method based on 3D convolution neural network | |
CN109214298B (en) | Asian female color value scoring model method based on deep convolutional network | |
CN103578098B (en) | Method and device for extracting commodity body in commodity picture | |
CN106845475A (en) | Natural scene character detecting method based on connected domain | |
CN105894506B (en) | A kind of facial image fuzziness computational methods and device | |
CN109635636A (en) | The pedestrian that blocking characteristic based on attributive character and weighting blends recognition methods again | |
CN106446073A (en) | User grade division method and apparatus | |
CN106951869A (en) | A kind of live body verification method and equipment | |
CN107590491A (en) | A kind of image processing method and device | |
CN109191428A (en) | Full-reference image quality evaluating method based on masking textural characteristics | |
CN110245550A (en) | A kind of face noise data collection CNN training method based on overall cosine distribution | |
CN103577993B (en) | Color choosing method and device | |
CN103440494A (en) | Horrible image identification method and system based on visual significance analyses | |
CN110188600B (en) | Drawing evaluation method, system and storage medium | |
CN107526810A (en) | Establish method and device, methods of exhibiting and the device of clicking rate prediction model | |
CN109829924A (en) | A kind of image quality evaluating method based on body feature analysis |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170926 |