CN108449596A - A kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level - Google Patents
A kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level Download PDFInfo
- Publication number
- CN108449596A CN108449596A CN201810342793.3A CN201810342793A CN108449596A CN 108449596 A CN108449596 A CN 108449596A CN 201810342793 A CN201810342793 A CN 201810342793A CN 108449596 A CN108449596 A CN 108449596A
- Authority
- CN
- China
- Prior art keywords
- feature
- stereo
- parallax
- indicate
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N17/00—Diagnosis, testing or measuring for television systems or their details
- H04N17/004—Diagnosis, testing or measuring for television systems or their details for digital television systems
-
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
- H04N2013/0074—Stereoscopic image analysis
- H04N2013/0081—Depth or disparity estimation from stereoscopic image signals
Abstract
The present invention relates to a kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level, include the following steps:Step S1:The aesthetic features of left and right view and left and right view aesthetic consistency feature are extracted to every width stereo-picture in training image set and image collection to be predicted, obtain aesthetic features collection F1;Step S2:Comfort feature is extracted to every width stereo-picture in training image set and image collection to be predicted, obtains comfort feature collection F2;Step S3:To all images in training image set, in conjunction with aesthetic features collection F1 and comfort feature collection F2, as machine learning feature set T1, training obtains stereo image quality assessment models;Step S4:Every image to be predicted is assessed using trained Evaluation Model on Quality, obtains the final mass assessment score of all images to be predicted.This method is conducive to improve the consistency of assessment result and user's subjective scores.
Description
Technical field
The present invention relates to image and video processing and computer vision field, especially a kind of fusion aesthetics and comfort level
3D stereo image quality appraisal procedures.
Background technology
The quality evaluation algorithm of image, which is divided into, reference, half reference and the quality evaluation without reference.Wherein, the matter of no reference
Amount assessment is intended to, by using the modes such as feature possessed by image itself, comment to carry out quality without corresponding reference picture
Estimate.
One 3D stereo-picture is made of left and right view, this also allows for 3D stereo-pictures both depth with binocular image
Feature also includes the feature of monocular image.In actual life, the aesthetic features of monocular image are most intuitively experienced as to user.
Niu et al. is mentioned when evaluating the frame picture of professional video, and requirement of the professional photographer for picture is less dominant hue number
Mesh, moderate color saturation and smooth brightness change range etc..
It is compared with normal image, the three-dimensional sense of binocular image generates parallax from left and right view, in human eye view
Imaging on film has differences, and this difference spatially is seemingly just the stereoscopic landscape one that human eye is seen in real world
Sample.Lambooij et al. proposes that fuzzy, color space mismatch under excessive parallax value, some undernatured states etc. is all
It can cause these visual discomforts.In parallax, it is divided into be imaged negative parallax before screen and being imaged on facing after screen
Difference and parallax free plane.Wherein, negative parallax is an important factor in order for causing visual comfort to be experienced.Shao et al.
In the visual comfort experience of research stereo-picture, in order to especially analyze the shadow generated in larger parallax value in stereo-picture
Situation is rung, by before minimum and maximum 10% parallax data independent analysis.As far as we know, apart from parallax free plane
Within a certain range, there are an euphorosia region, the parallax value in the region meets the Natural regulation of human eye ball, also not
Can there are problems that the various aspects such as visual imaging conflict.But it all can inevitably lead to parallax in the process of shooting and post-production
Numerical value is except this comfort zone range, this makes the disparity range of stereo-picture, and there are probabilistic variations.
Invention content
The purpose of the present invention is to provide a kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level, the party
Method is conducive to improve the consistency of assessment result and user's subjective scores.
To achieve the above object, the technical scheme is that:A kind of 3D stereogram image qualities of fusion aesthetics and comfort level
Appraisal procedure is measured, is included the following steps:
Step S1:It is three-dimensional to every width in the two set to input training image set, image collection to be predicted and user
The subjective quality assessment score of image;Left to every width stereo-picture extraction in training image set and image collection to be predicted,
The aesthetic features of right view and left and right view aesthetic consistency feature obtain aesthetic features collection F1;
Step S2:Comfort feature is extracted to every width stereo-picture in training image set and image collection to be predicted,
Obtain comfort feature collection F2;
Step S3:It will in conjunction with aesthetic features collection F1 and comfort feature collection F2 to all images in training image set
It obtains stereo image quality assessment models as machine learning feature set T1, training;
Step S4:Every image to be predicted is assessed using trained Evaluation Model on Quality, obtains all images to be predicted
Final mass assess score.
Further, in the step S1, the aesthetic features and left and right view to the left and right view of three-dimensional image zooming-out are beautiful
Consistency feature is learned, aesthetic features collection F1 is obtained, includes the following steps:
Step S11:To the left and right view of three-dimensional image zooming-out in dominant hue number, color saturation and brightness range three
The aesthetic features of a aspect, calculation formula are:
Lt=L0(β*W*H)-L0((1-β)*W*H+1)
Wherein, t indicates that the left view or right view of stereo-picture, t=L indicate that left view, t=R indicate right view;HtTable
Show the dominant hue number of t views;T views are converted into hsv color space from RGB color, by the channels H numerical value n deciles
Color histogram indicate,Indicate that the color histogram numerical value of jth decile in t views, m are maximum in color histogram
The numerical value of decile;α is the parameter of setting;It indicates to solve dominant hue set, when the color histogram of a certain decile
Figure numerical value is more than α m, i.e. when the α multiples of m, that is, thinks that the corresponding color of the decile belongs to dominant hue set;Count () indicates meter
Calculate element number therein;StIndicate the color saturation mean value of t views;W, H indicates the width and height of image respectively, S (i,
J) numerical value of pixel (i, j) channel S in hsv color space is indicated;LtIndicate that t views L in CIE LUV color spaces is logical
The numerical value in road, L0It indicates the obtained sequence of the ascending sequence sequence of the channels the L numerical value of all pixels point in t views, figure
It is the parameter of setting, β ∈ [0,1] that the brightness range of picture, which takes the part numberical range for accounting for β times of the sequence centre, β,;
Step S12:Consistency feature of the left and right view of stereo-picture on aesthetic features is calculated, calculation formula is:
Hc=| HL-HR|, Sc=| SL-SR|, Lc=| LL-LR|
Wherein, Hc, Sc, Lc indicate consistency feature of the left and right view of stereo-picture on aesthetic features respectively, successively
For:Dominant hue number consistency, color saturation consistency and brightness range consistency;In addition, using color contrast is based on
Spend the solid colour between similarity and the image quality measure method CSVD calculating left and right views of stereo-picture of color data error
Property FCSVD;
Combining step S11-S12 obtains aesthetic features collection F1={ Ht,St,Lt,Hc,Sc,Lc,FCSVD}。
Further, in the step S2, to every width stereo-picture in training image set and image collection to be predicted
Comfort feature is extracted, comfort feature collection F2 is obtained, includes the following steps:
Step S21:The horizontal and vertical disparity map of stereo-picture is calculated using SIFT Flow dense Stereo Matching algorithms;
On the basis of the disparity map arrived, comfort feature is calculated from positive parallax, negative parallax, parallax mean value, parallax variance many aspects;Meter
It is as follows to calculate formula:
Wherein,Horizontal positive parallax, horizontal negative parallax, vertical positive parallax are indicated respectively and are vertically born
Parallax;W and H indicates the width and height of image respectively;Vx(i, j) and Vy(i, j) indicates stereo-picture at (i, j) respectively
Horizontal and vertical parallax value;N (Ω+) and N (Ω -) indicates the pixel in positive parallax set omega+and negative parallax set omega-respectively
Number;Dd indicates disparity range, drelativeIndicate the relative depth of parallax;On the basis of calculating parallax mean value, calculate every
The corresponding variance of mean value, calculation formula are as follows:
Wherein,The variance of horizontal positive parallax, horizontal negative parallax are indicated respectively
The variance of variance, the variance of vertical positive parallax and vertical negative parallax;Std (z) indicates to solve the side of all elements in set z
Difference;
Step S22:Calculate edge parallax feature;The parallax of preceding t% is in calculated level positive parallax and horizontal negative parallax
Value, calculation formula are as follows:
Wherein, dmaxIndicate that absolute value is more than the positive parallax mean value of preceding percent t, dminIndicate that absolute value is more than preceding percent
The negative parallax mean value of t,Indicate that absolute value faces difference set in preceding t% respectivelyWith absolute value in preceding t%
Negative parallax setIn pixel number;
Step S23:Calculate spatial frequency correlated characteristic;The spatial frequency features for calculating separately left and right view, then take two
Person's mean value characterizes the spatial frequency of stereo-picture, and calculation formula is as follows:
Wherein, fl, fr indicate the spatial frequency features of left and right view, SB respectivelyl(i,j)、SBr(i, j) indicate respectively it is left,
Right view utilizes the numerical value of the calculated each pixel of sobel edge detection operators at (i, j);σ1、σ2、σ3F is indicated respectively
Feature is contacted with what parallax feature was established;
Step S24:Computation vision comfort zone correlated characteristic;Calculation formula is as follows:
Wherein, γ+Indicate the adjustable threshold value for imaging in the euphorosia area before screen of retina, γ-Indicate that retina can
The threshold value for imaging in the euphorosia area after screen is adjusted, ρ indicates pupil diameter;S indicates that bulbous length, v indicate viewing distance;
Work as γ+And γ-When beyond human eye control range, stereo-picture will produce it is fuzzy, increase the feeling of fatigue of viewer by;
Combining step S21-S24 obtains comfort feature collection, as follows:
Further, it in the step S3, to all images in training image set, in conjunction with aesthetic features collection F1 and relaxes
Appropriate feature set F2, as machine learning feature set T1, training obtains stereo image quality assessment models, specific method
For:
In Fusion training data acquisition system the feature set F1 and F2 of all images and by user to institute in training image set
Have a tally set L1 that the subjective quality assessment score of image obtains, form training image set feature set T1={ F1, F2 } and
Tally set L1;It is trained by feature set T1 and tally set L1 using random forest homing method, obtains stereo image quality
Assessment models M.
Further, in the step S4, the feature of all images in data acquisition system to be predicted is merged, figure to be predicted is formed
The feature set T2={ F1, F2 } that image set closes is calculated using the stereo image quality assessment models that training obtains in step S3
The final mass of all images to be predicted assesses score.
Compared to the prior art, the beneficial effects of the invention are as follows:The present invention has merged the aesthetics for influencing stereo image quality
Feature and comfort feature propose a kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level, it is vertical to solve 3D
The shortcomings that body image carrys out quality of evaluation from single angle, the 3D stereo image qualities score that this method is assessed can be with user
Subjective scores keep higher consistency, can be used for image quality measure, the fields such as image or visual classification.
Description of the drawings
Fig. 1 is the flow diagram of the method for the present invention.
Fig. 2 is the implementation flow chart of holistic approach in the embodiment of the present invention.
Specific implementation mode
Below in conjunction with the accompanying drawings and specific embodiment, the present invention is described in further details.
The present invention provides a kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level, such as Fig. 1 and Fig. 2 institutes
Show, includes the following steps:
Step S1:It is three-dimensional to every width in the two set to input training image set, image collection to be predicted and user
The subjective quality assessment score of image.Left to every width stereo-picture extraction in training image set and image collection to be predicted,
The aesthetic features of right view and left and right view aesthetic consistency feature obtain aesthetic features collection F1.Specifically include following steps:
Step S11:To the left and right view of three-dimensional image zooming-out in dominant hue number, color saturation and brightness range three
The aesthetic features of a aspect, calculation formula are:
Lt=L0(β*W*H)-L0((1-β)*W*H+1)
Wherein, t indicates that the left view or right view of stereo-picture, t=L indicate that left view, t=R indicate right view;HtTable
Show the dominant hue number of t views;T views are converted into hsv color space from RGB color, by the channels H numerical value n deciles
Color histogram indicate (n takes 20 in the present embodiment),Indicating the color histogram numerical value of jth decile in t views, m is
The numerical value of maximum decile in color histogram;α is the parameter of setting;It indicates to solve dominant hue set, when certain
The color histogram numerical value of one decile is more than α m, i.e. when the α multiples of m, that is, thinks that the corresponding color of the decile belongs to mass-tone and assembles
It closes;Count () indicates to calculate element number therein;StIndicate the color saturation mean value of t views;W, H indicates image respectively
Width and height, S (i, j) indicate pixel (i, j) in hsv color space channel S numerical value;Brightness is influence diagram image quality
One factor of amount, LtIndicate the numerical value in t views channels L in CIE LUV color spaces, L0It indicates all pixels in t views
The obtained sequence of the ascending sequence sequence of the channels L numerical value of point, the brightness range of image, which takes, accounts among the sequence β times
Part numberical range, β are the parameter of setting, β ∈ [0,1];In the present embodiment, β takes 90%;
Step S12:Consistency feature of the left and right view of stereo-picture on aesthetic features is calculated, calculation formula is:
Hc=| HL-HR|, Sc=| SL-SR|, Lc=| LL-LR|
Wherein, Hc, Sc, Lc indicate consistency feature of the left and right view of stereo-picture on aesthetic features respectively, successively
For:Dominant hue number consistency, color saturation consistency and brightness range consistency;U.S. between left and right two view
When feature difference is excessive, the quality of stereo-picture can be had an impact, therefore the present invention considers left and right view aesthstic special
Consistency feature in sign;In addition, utilizing the image quality measure method based on color contrast similarity and color data error
CSVD calculates the colour consistency F between the left and right view of stereo-pictureCSVD;
Combining step S11-S12 obtains aesthetic features collection F1={ Ht,St,Lt,Hc,Sc,Lc,FCSVD}。
Step S2:Comfort feature is extracted to every width stereo-picture in training image set and image collection to be predicted,
Obtain comfort feature collection F2.Specifically include following steps:
Step S21:Pixel exists between calculating stereo-picture middle left and right view using SIFT Flow dense Stereo Matching algorithms
Situation of movement on horizontally and vertically obtains the horizontal and vertical disparity map of stereo-picture according to obtained result;
On the basis of obtained disparity map, comfort level is calculated from many aspects such as positive parallax, negative parallax, parallax mean value, parallax variances
Feature;Calculation formula is as follows:
Wherein,Horizontal positive parallax, horizontal negative parallax, vertical positive parallax are indicated respectively and are vertically born
Parallax;W and H indicates the width and height of image respectively;Vx(i, j) and Vy(i, j) indicates stereo-picture at (i, j) respectively
Horizontal and vertical parallax value;N(Ω+) and N (Ω-) positive parallax set omega is indicated respectively+With negative parallax set omega-In pixel
Number;Dd indicates disparity range, drelativeIndicate the relative depth of parallax;On the basis of calculating parallax mean value, calculate every
The corresponding variance of mean value, calculation formula are as follows:
Wherein,The variance of horizontal positive parallax, the side of horizontal negative parallax are indicated respectively
The variance of poor, vertical positive parallax and the variance of vertical negative parallax;Std (z) indicates to solve the variance of all elements in set z;
Step S22:Calculate edge parallax feature;Since excessive parallax value can make human eye generation uncomfortable, those are in
The pixel number of larger parallax is generally few, but influence of this partial pixel point to viewing experience is that can not ignore,
So being in the parallax value of preceding t% in calculated level positive parallax of the present invention and horizontal negative parallax, calculation formula is as follows:
Wherein, dmaxIndicate that absolute value is more than the positive parallax mean value of preceding percent t, dminIndicate that absolute value is more than preceding percent
The negative parallax mean value of t,Indicate that absolute value faces difference set in preceding t% respectivelyWith absolute value in preceding t%
Negative parallax setIn pixel number;
Step S23:Calculate spatial frequency correlated characteristic;In stereo-picture, spatial frequency has the fusion of binocular image
It has a major impact;Research shows that the limiting value of the more high corresponding binocular fusion of spatial frequency is smaller, so higher spatial frequency
Binocular fusion of the stereo-picture in human eye system can be limited;Binocular fusion is to assess the key factor of stereo image quality;And
Spatial frequency is to influence an important feature of binocular fusion, therefore calculate separately the spatial frequency features of left and right view, then
The two mean value is taken to characterize the spatial frequency of stereo-picture, calculation formula is as follows:
Wherein, fl, fr indicate the spatial frequency features of left and right view, SB respectivelyl(i,j)、SBr(i, j) indicate respectively it is left,
Right view utilizes the numerical value of the calculated each pixel of sobel edge detection operators at (i, j);σ1、σ2、σ3F is indicated respectively
Feature is contacted with what parallax feature was established;
Step S24:Computation vision comfort zone correlated characteristic;Judge stereo-picture by whether being in after retina image-forming
Euphorosia area, be assess stereo-picture whether be high quality graphic a key factor;Utilize viewing distance, retina half
The parameters such as diameter, euphorosia area correlated characteristic calculation formula are as follows:
Wherein, γ+Indicate the adjustable threshold value for imaging in the euphorosia area before screen of retina, γ-Indicate that retina can
The threshold value for imaging in the euphorosia area after screen is adjusted, ρ indicates pupil diameter;S indicates that bulbous length, v indicate viewing distance;
When the numerical value exceeds human eye control range, stereo-picture will produce it is fuzzy, increase the feeling of fatigue of viewer by;
Combining step S21-S24 obtains comfort feature collection, as follows:
Step S3:It will in conjunction with aesthetic features collection F1 and comfort feature collection F2 to all images in training image set
It obtains stereo image quality assessment models as machine learning feature set T1, training.Specific method is:
In Fusion training data acquisition system the feature set F1 and F2 of all images and by user to institute in training image set
Have a tally set L1 that the subjective quality assessment score of image obtains, form training image set feature set T1={ F1, F2 } and
Tally set L1;It is trained by feature set T1 and tally set L1 using random forest homing method, obtains stereo image quality
Assessment models M.
Step S4:The feature for merging all images in data acquisition system to be predicted forms the feature set of image collection to be predicted
T2={ F1, F2 } assesses every image to be predicted using trained stereo image quality assessment models, obtains all to be predicted
The final mass of image assesses score.
The above are preferred embodiments of the present invention, all any changes made according to the technical solution of the present invention, and generated function is made
When with range without departing from technical solution of the present invention, all belong to the scope of protection of the present invention.
Claims (5)
1. a kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level, which is characterized in that include the following steps:
Step S1:Training image set, image collection to be predicted and user are inputted to every width stereo-picture in the two set
Subjective quality assessment score;It is left and right to every width stereo-picture extraction in training image set and image collection to be predicted to regard
The aesthetic features of figure and left and right view aesthetic consistency feature obtain aesthetic features collection F1;
Step S2:Comfort feature is extracted to every width stereo-picture in training image set and image collection to be predicted, is obtained
Comfort feature collection F2;
Step S3:All images in training image set are made in conjunction with aesthetic features collection F1 and comfort feature collection F2
For machine learning feature set T1, training obtains stereo image quality assessment models;
Step S4:Every image to be predicted is assessed using trained Evaluation Model on Quality, obtains all images to be predicted most
Whole quality evaluation score.
2. the 3D stereo image quality appraisal procedures of a kind of fusion aesthetics and comfort level according to claim 1, feature
It is, in the step S1, the aesthetic features to the left and right view of three-dimensional image zooming-out and left and right view aesthetic consistency feature,
Aesthetic features collection F1 is obtained, is included the following steps:
Step S11:To the left and right view of three-dimensional image zooming-out in three dominant hue number, color saturation and brightness range sides
The aesthetic features in face, calculation formula are:
Lt=L0(β*W*H)-L0((1-β)*W*H+1)
Wherein, t indicates that the left view or right view of stereo-picture, t=L indicate that left view, t=R indicate right view;HtIndicate that t is regarded
The dominant hue number of figure;T views are converted into hsv color space from RGB color, by the channels the H numerical value color of n deciles
Histogram expression,Indicate that the color histogram numerical value of jth decile in t views, m are maximum deciles in color histogram
Numerical value;α is the parameter of setting;It indicates to solve dominant hue set, when the color histogram numerical value of a certain decile
More than α m, i.e. when the α multiples of m, that is, think that the corresponding color of the decile belongs to dominant hue set;Count () indicates to calculate wherein
Element number;StIndicate the color saturation mean value of t views;W, H indicates that the width and height of image, S (i, j) indicate respectively
The numerical value of pixel (i, j) channel S in hsv color space;LtIndicate the number in t views channels L in CIE LUV color spaces
Value, L0Indicate to sort the ascending sequence of the channels the L numerical value of all pixels point in t views obtained sequence, image it is bright
It is the parameter of setting, β ∈ [0,1] that degree range, which takes the part numberical range for accounting for β times of the sequence centre, β,;
Step S12:Consistency feature of the left and right view of stereo-picture on aesthetic features is calculated, calculation formula is:
Hc=| HL-HR|, Sc=| SL-SR|, Lc=| LL-LR|
Wherein, Hc, Sc, Lc indicate consistency feature of the left and right view of stereo-picture on aesthetic features respectively, are followed successively by:It is main
Tone number consistency, color saturation consistency and brightness range consistency;In addition, using similar based on color contrast
The image quality measure method CSVD of degree and color data error calculates the colour consistency between the left and right view of stereo-picture
FCSVD;
Combining step S11-S12 obtains aesthetic features collection F1={ Ht,St,Lt,Hc,Sc,Lc,FCSVD}。
3. the 3D stereo image quality appraisal procedures of a kind of fusion aesthetics and comfort level according to claim 1, feature
It is, it is special to every width stereo-picture extraction comfort level in training image set and image collection to be predicted in the step S2
Sign obtains comfort feature collection F2, includes the following steps:
Step S21:The horizontal and vertical disparity map of stereo-picture is calculated using SIFT Flow dense Stereo Matching algorithms;What is obtained
On the basis of disparity map, comfort feature is calculated from positive parallax, negative parallax, parallax mean value, parallax variance many aspects;It calculates public
Formula is as follows:
Dd=max { Vx(i,j)}-min{Vx(i,j)},
Wherein,Horizontal positive parallax, horizontal negative parallax, vertical positive parallax and vertical negative parallax are indicated respectively;
W and H indicates the width and height of image respectively;Vx(i, j) and Vy(i, j) respectively indicate stereo-picture at (i, j) level and
Vertical parallax value;N(Ω+) and N (Ω-) positive parallax set omega is indicated respectively+With negative parallax set omega-In pixel number;dd
Indicate disparity range, drelativeIndicate the relative depth of parallax;On the basis of calculating parallax mean value, calculates every mean value and correspond to
Variance, calculation formula is as follows:
Wherein,Indicate respectively the variance of horizontal positive parallax, the variance of horizontal negative parallax,
The vertical variance of positive parallax and the variance of vertical negative parallax;Std (z) indicates to solve the variance of all elements in set z;
Step S22:Calculate edge parallax feature;The parallax value of preceding t%, meter are in calculated level positive parallax and horizontal negative parallax
It is as follows to calculate formula:
Wherein, dmaxIndicate that absolute value is more than the positive parallax mean value of preceding percent t, dminIndicate absolute value more than preceding percent t's
Negative parallax mean value,Indicate that absolute value faces difference set in preceding t% respectivelyIt bears and regards in preceding t% with absolute value
Difference setIn pixel number;
Step S23:Calculate spatial frequency correlated characteristic;The spatial frequency features of left and right view are calculated separately, are then taken both
Value characterizes the spatial frequency of stereo-picture, and calculation formula is as follows:
F=(fl+fr)/2
Wherein, fl, fr indicate the spatial frequency features of left and right view, SB respectivelyl(i,j)、SBr(i, j) indicates left and right and regards respectively
Figure utilizes the numerical value of the calculated each pixel of sobel edge detection operators at (i, j);σ1、σ2、σ3It indicates f respectively and regards
The contact feature that poor feature is established;
Step S24:Computation vision comfort zone correlated characteristic;Calculation formula is as follows:
Wherein, γ+Indicate the adjustable threshold value for imaging in the euphorosia area before screen of retina, γ-Indicate that retina is adjustable
The threshold value in the euphorosia area after screen is imaged in, ρ indicates pupil diameter;S indicates that bulbous length, v indicate viewing distance;When
γ+And γ-When beyond human eye control range, stereo-picture will produce it is fuzzy, increase the feeling of fatigue of viewer by;
Combining step S21-S24 obtains comfort feature collection, as follows:
4. the 3D stereo image quality appraisal procedures of a kind of fusion aesthetics and comfort level according to claim 1, feature
It is, in the step S3, to all images in training image set, in conjunction with aesthetic features collection F1 and comfort feature collection
F2, as machine learning feature set T1, training obtains stereo image quality assessment models, and specific method is:
In Fusion training data acquisition system the feature set F1 and F2 of all images and by user to all figures in training image set
The tally set L1 that the subjective quality assessment score of picture obtains forms the feature set T1={ F1, F2 } and label of training image set
Collect L1;It is trained by feature set T1 and tally set L1 using random forest homing method, obtains stereo image quality assessment
Model M.
5. the 3D stereo image quality appraisal procedures of a kind of fusion aesthetics and comfort level according to claim 1, feature
It is, in the step S4, merges the feature of all images in data acquisition system to be predicted, form the feature of image collection to be predicted
Collect T2={ F1, F2 }, using the stereo image quality assessment models that training obtains in step S3, all figures to be predicted are calculated
The final mass of picture assesses score.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810342793.3A CN108449596B (en) | 2018-04-17 | 2018-04-17 | 3D stereoscopic image quality evaluation method integrating aesthetics and comfort |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810342793.3A CN108449596B (en) | 2018-04-17 | 2018-04-17 | 3D stereoscopic image quality evaluation method integrating aesthetics and comfort |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108449596A true CN108449596A (en) | 2018-08-24 |
CN108449596B CN108449596B (en) | 2020-09-01 |
Family
ID=63200047
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810342793.3A Expired - Fee Related CN108449596B (en) | 2018-04-17 | 2018-04-17 | 3D stereoscopic image quality evaluation method integrating aesthetics and comfort |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108449596B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109523590A (en) * | 2018-10-22 | 2019-03-26 | 福州大学 | A kind of 3D rendering depth information visual comfort appraisal procedure based on sample |
CN109584326A (en) * | 2018-11-07 | 2019-04-05 | 厦门欢乐逛科技股份有限公司 | Poster Automatic color matching method, medium, computer equipment and device |
CN109801256A (en) * | 2018-12-15 | 2019-05-24 | 华南理工大学 | A kind of image aesthetic quality appraisal procedure based on area-of-interest and global characteristics |
CN109815965A (en) * | 2019-02-13 | 2019-05-28 | 腾讯科技(深圳)有限公司 | A kind of image filtering method, device and storage medium |
CN110533097A (en) * | 2019-08-27 | 2019-12-03 | 腾讯科技(深圳)有限公司 | A kind of image definition recognition methods, device, electronic equipment and storage medium |
CN111046953A (en) * | 2019-12-12 | 2020-04-21 | 电子科技大学 | Image evaluation method based on similarity comparison |
CN112738501A (en) * | 2020-12-29 | 2021-04-30 | 杭州电子科技大学 | Three-dimensional image comfort level testing method |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574885A (en) * | 2016-01-13 | 2016-05-11 | 福州大学 | Machine-learning-based full-reference image quality assessment method integrated with visual characteristics |
CN105678775A (en) * | 2016-01-13 | 2016-06-15 | 福州大学 | Color correction assessment method based on machine learning |
CN107396095A (en) * | 2017-08-28 | 2017-11-24 | 方玉明 | One kind is without with reference to three-dimensional image quality evaluation method |
CN107578404A (en) * | 2017-08-22 | 2018-01-12 | 浙江大学 | The complete of view-based access control model notable feature extraction refers to objective evaluation method for quality of stereo images |
-
2018
- 2018-04-17 CN CN201810342793.3A patent/CN108449596B/en not_active Expired - Fee Related
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105574885A (en) * | 2016-01-13 | 2016-05-11 | 福州大学 | Machine-learning-based full-reference image quality assessment method integrated with visual characteristics |
CN105678775A (en) * | 2016-01-13 | 2016-06-15 | 福州大学 | Color correction assessment method based on machine learning |
CN107578404A (en) * | 2017-08-22 | 2018-01-12 | 浙江大学 | The complete of view-based access control model notable feature extraction refers to objective evaluation method for quality of stereo images |
CN107396095A (en) * | 2017-08-28 | 2017-11-24 | 方玉明 | One kind is without with reference to three-dimensional image quality evaluation method |
Non-Patent Citations (2)
Title |
---|
JIAMEI ZHAN,YUZHEN NIU,等: "Learning from multi metrics for stereoscopic 3D image quality assessment", 《2016 INTERNATIONAL CONFERENCE ON 3D IMAGING (IC3D)》 * |
YUZHEN NIU,FENG LIU,等: "Aesthetics-Based Stereoscopic Photo Cropping for Heterogeneous Displays", 《IEEE TRANSACTIONS ON MULTIMEDIA》 * |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109523590A (en) * | 2018-10-22 | 2019-03-26 | 福州大学 | A kind of 3D rendering depth information visual comfort appraisal procedure based on sample |
CN109523590B (en) * | 2018-10-22 | 2021-05-18 | 福州大学 | 3D image depth information visual comfort evaluation method based on sample |
CN109584326A (en) * | 2018-11-07 | 2019-04-05 | 厦门欢乐逛科技股份有限公司 | Poster Automatic color matching method, medium, computer equipment and device |
CN109801256A (en) * | 2018-12-15 | 2019-05-24 | 华南理工大学 | A kind of image aesthetic quality appraisal procedure based on area-of-interest and global characteristics |
CN109801256B (en) * | 2018-12-15 | 2023-05-26 | 华南理工大学 | Image aesthetic quality assessment method based on region of interest and global features |
CN109815965A (en) * | 2019-02-13 | 2019-05-28 | 腾讯科技(深圳)有限公司 | A kind of image filtering method, device and storage medium |
CN109815965B (en) * | 2019-02-13 | 2021-07-06 | 腾讯科技(深圳)有限公司 | Image filtering method and device and storage medium |
CN110533097B (en) * | 2019-08-27 | 2023-01-06 | 腾讯科技(深圳)有限公司 | Image definition recognition method and device, electronic equipment and storage medium |
CN110533097A (en) * | 2019-08-27 | 2019-12-03 | 腾讯科技(深圳)有限公司 | A kind of image definition recognition methods, device, electronic equipment and storage medium |
CN111046953A (en) * | 2019-12-12 | 2020-04-21 | 电子科技大学 | Image evaluation method based on similarity comparison |
CN111046953B (en) * | 2019-12-12 | 2022-06-21 | 电子科技大学 | Image evaluation method based on similarity comparison |
CN112738501A (en) * | 2020-12-29 | 2021-04-30 | 杭州电子科技大学 | Three-dimensional image comfort level testing method |
CN112738501B (en) * | 2020-12-29 | 2022-05-17 | 杭州电子科技大学 | Three-dimensional image comfort level testing method |
Also Published As
Publication number | Publication date |
---|---|
CN108449596B (en) | 2020-09-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108449596A (en) | A kind of 3D stereo image quality appraisal procedures of fusion aesthetics and comfort level | |
CN103763552B (en) | Stereoscopic image non-reference quality evaluation method based on visual perception characteristics | |
CN107578403B (en) | The stereo image quality evaluation method for instructing binocular view to merge based on gradient information | |
CN107181940B (en) | A kind of three-dimensional video-frequency Comfort Evaluation method and device | |
CN103986925B (en) | based on the stereoscopic video visual comfort evaluation method of luminance compensation | |
CN102750695A (en) | Machine learning-based stereoscopic image quality objective assessment method | |
CN105046708A (en) | Color correction objective assessment method consistent with subjective perception | |
CN104853185A (en) | Stereo video comfort evaluation method combining multiple parallaxes with motion | |
CN111354077B (en) | Binocular vision-based three-dimensional face reconstruction method | |
KR20110014067A (en) | Method and system for transformation of stereo content | |
TWI457853B (en) | Image processing method for providing depth information and image processing system using the same | |
CN108324247B (en) | Method and system for evaluating skin wrinkles at specified positions | |
CN106874830A (en) | A kind of visually impaired people's householder method based on RGB D cameras and recognition of face | |
CN111091046A (en) | User bad sitting posture correction system based on machine vision | |
CN108257165A (en) | Image solid matching method, binocular vision equipment | |
Chen et al. | Study of subject agreement on stereoscopic video quality | |
CN111008971A (en) | Aesthetic quality evaluation method of group photo image and real-time shooting guidance system | |
Yun et al. | The objective quality assessment of stereo image | |
CN207589060U (en) | A kind of naked-eye stereoscopic display device of combination visual fatigue detection | |
CN104469355B (en) | Based on the prediction of notable adaptive euphoropsia and the euphoropsia Enhancement Method of nonlinear mapping | |
CN106683072B (en) | 3D image comfort level quality evaluation method and system based on PUP image | |
CN107680076A (en) | Stereo-picture comfort level contrast range measuring method based on salient region | |
CN109167988B (en) | Stereo image visual comfort evaluation method based on D + W model and contrast | |
CN106993183A (en) | Comfortable brightness quantitative approach based on stereo-picture marking area | |
CN110060291B (en) | Three-dimensional apparent distance resolving method considering human factors |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200901 |
|
CF01 | Termination of patent right due to non-payment of annual fee |