CN105163111B - Method for evaluating visual comfort of three-dimensional image based on classification of scene modes - Google Patents
Method for evaluating visual comfort of three-dimensional image based on classification of scene modes Download PDFInfo
- Publication number
- CN105163111B CN105163111B CN201510571897.8A CN201510571897A CN105163111B CN 105163111 B CN105163111 B CN 105163111B CN 201510571897 A CN201510571897 A CN 201510571897A CN 105163111 B CN105163111 B CN 105163111B
- Authority
- CN
- China
- Prior art keywords
- stereo
- picture
- scene mode
- pixel
- screen
- 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.)
- Active
Links
Abstract
The invention discloses a method for evaluating the visual comfort of a three-dimensional image based on classification of scene modes. The method comprises the following steps: determining ten kinds of scene modes of the three-dimensional image according to depth positions, where a foreground object and a background area in a natural scene are; then, determining the scene mode, which each three-dimensional image belongs to, in a three-dimensional image database, and establishing a visual comfort evaluation mode in each scene mode; and firstly determining the scene mode, which the three-dimensional image to be evaluated belongs to, then, selecting the visual comfort evaluation mode in the scene mode, calculating a visual comfort evaluation predication value of the three-dimensional image to be evaluated, and finally, correcting the visual comfort evaluation predication value of the three-dimensional image to be evaluated. The method disclosed by the invention has the advantages that: perception of a human visual system to the visual comfort of the three-dimensional image can be precisely reflected through the visual comfort evaluation modes respectively established based on ten kinds of scenes; and thus, the consistency between an objective evaluation result and human eye subjective perception can be effectively improved.
Description
Technical field
The present invention relates to a kind of stereo-picture Quality of experience evaluation method, especially relate to a kind of scene mode that is based on and classify
Stereo image vision comfort level evaluation method.
Background technology
With the fast development of Stereoscopic Video Presentation technology and high-quality stereoscopic video content acquiring technology, three-dimensional video-frequency
Visual experience quality (qoe, quality of experience) is one of three-dimensional video-frequency system design major issue, and
Visual comfort (vc, visual comfort) is one of key factor of visual experience quality of impact three-dimensional video-frequency.At present,
Research contents distortion is concentrated mainly on for visual experience quality to the visual experience matter quantifier elimination of three-dimensional video-frequency/stereo-picture
Impact, this respect created many achievements in research.The research of stereo image vision comfort level is in stereoscopic image content not
The physiological discomfort causing under the premise of distortion, the visual comfort objective evaluation model obtaining is for raising beholder's
Visual experience quality, the making instructing 3d content and later stage process and have a very important role.
Existing stereo image vision comfort level evaluation method is mainly passed through to extract the left and right visual point image of stereo-picture
And the feature of anaglyph, such as parallax, gradient etc., then set up using the algorithm of method statistically or machine learning
Play objective models.But at present to causing the various factors of physiologically discomfort to make clear of not yet in the content of stereo-picture
Thorough, lead to the uniformity between objective evaluation result and human eye subjective perception poor.Relax for obtaining the preferable vision of evaluation effect
Appropriate evaluation method, needs all kinds of natural scenes are finely divided, and on this basis the various characteristics of image of Accurate Analysis to regarding
Feel the influence degree of comfort level.
Content of the invention
The technical problem to be solved is to provide a kind of stereo-picture euphorosia based on scene mode classification
Degree evaluation method, it can effectively improve the uniformity between objective evaluation result and human eye subjective perception.
The technical scheme that present invention solution above-mentioned technical problem is adopted is: a kind of stereogram based on scene mode classification
As visual comfort evaluation method it is characterised in that comprising the following steps:
1. 10 kinds of scene modes of the natural scene shown by three-dimensional display are determined, the 1st kind of scene mode is stereogram
Foreground target in the right viewpoint parallax gray level image of picture convex in screen and be in non-comfort zone, background region recessed in screen and place
In non-comfort zone;2nd kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and
Be in non-comfort zone, background region convex in screen and be in non-comfort zone;3rd kind of scene mode is that the right viewpoint of stereo-picture regards
Foreground target in difference gray level image recessed in screen and be in non-comfort zone, background region recessed in screen and be in non-comfort zone;
4th kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in non-comfortable
Area, background region recessed in screen and be in comfort zone;5th kind of scene mode is in the right viewpoint parallax gray level image of stereo-picture
Foreground target convex in screen and be in non-comfort zone, background region convex in screen and be in comfort zone;6th kind of scene mode be
Foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in comfort zone, background region recessed in screen
And it is in non-comfort zone;7th kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture recessed in screen
Curtain and be in comfort zone, background region recessed in screen and be in non-comfort zone;8th kind of scene mode is the right viewpoint of stereo-picture
Foreground target in parallax gray level image convex in screen and be in comfort zone, background region recessed in screen and be in comfort zone;9th
Kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in comfort zone, background
Region convex in screen and be in comfort zone;10th kind of scene mode is the prospect in the right viewpoint parallax gray level image of stereo-picture
Target recessed in screen and be in comfort zone, background region recessed in screen and be in comfort zone;
2. every width stereo-picture is selected to have a stereoscopic image data storehouse of the mean subjective scoring average of visual comfort;
Then the parallax according to the foreground target in the right viewpoint parallax gray level image of the every width stereo-picture in stereoscopic image data storehouse
Angle and the parallactic angle of background region, determine the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse;Further according to
Before belonging in stereoscopic image data storehouse in the respective right viewpoint parallax gray level image of all stereo-pictures of every kind of scene mode
The parallactic angle of scape target and the parallactic angle of background region, set up the visual comfort evaluation model under every kind of scene mode, by n-th
The visual comfort evaluation model planted under scene mode is described as follows: smmon=(4.2028-u (qn))-v(qn)×da(qn)+
0.1912×ln(wa)-0.0208×da(qn)×ln(wa), wherein, 1≤n≤10, smmonRepresent under n scene mode
The output of visual comfort evaluation model, u (qn) and v (qn) it is constant, da(qn) represent that a width to be entered belongs to n
The global disparity angle of the right viewpoint parallax gray level image of the stereo-picture to be evaluated of scene mode, da(qn)=qn×|fa|+
(1-qn)×|ba|, qnRepresent the weight under n scene mode, faThe width representing to be entered belongs to n scene mode
The parallactic angle of the foreground target in the right viewpoint parallax gray level image of stereo-picture to be evaluated, baThe width representing to be entered belongs to
The parallactic angle of the background region in the right viewpoint parallax gray level image of the stereo-picture to be evaluated of n scene mode, wa
Before in the right viewpoint parallax gray level image of stereo-picture to be evaluated that the width representing to be entered belongs to n scene mode
The width angle of scape target, symbol " | | " it is the symbol that takes absolute value;
3. the right viewpoint parallax gray level image of stereo-picture to be evaluated is designated as { dr(x, y) }, wherein, 1≤x≤w, 1
≤ y≤h, w represent the width of stereo-picture to be evaluated, its with stereoscopic image data storehouse in every width stereo-picture width one
Cause, h represents the height of stereo-picture to be evaluated, its with stereoscopic image data storehouse in every width stereo-picture highly consistent,
dr(x, y) represents { dr(x, y) } in coordinate position be (x, y) pixel pixel value;Then adopt and step 2. middle identical
Mode, determines the scene mode belonging to stereo-picture to be evaluated;Further according to the scene mode belonging to stereo-picture to be evaluated,
Choose the visual comfort evaluation model under this scene mode;Then according to the visual comfort evaluation model under this scene mode
And { dr(x, y) } in the parallactic angle of foreground target, { dr(x, y) } in the parallactic angle of background region and { dr(x, y) } in before
The width angle of scape target, calculates the visual comfort evaluation and foreca value of stereo-picture to be evaluated, is designated as smmo it is assumed that to be evaluated
Stereo-picture belong to n scene mode, then smmo=(4.2028-u (qn))-v(qn)×da'(qn)+0.1912×ln
(wa')-0.0208×da'(qn)×ln(wa'), wherein, da'(qn) represent { dr(x, y) } global disparity angle, da'(qn)=qn
×|fa'|+(1-qn)×|ba' |, fa' represent { dr(x, y) } in foreground target parallactic angle, ba' represent { dr(x, y) } in
The parallactic angle of background region, wa' represent { dr(x, y) } in foreground target width angle;
4. visual comfort evaluation and foreca value smmo of stereo-picture to be evaluated is modified, by solid to be evaluated
The revised visual comfort evaluation and foreca value of image is designated as smm,Wherein, max () is to take max function, p
The tortuosity attenuation coefficient of the visual comfort evaluation model under the expression scene mode belonging to stereo-picture to be evaluated, sr'
Represent { dr(x, y) } in foreground target average prospect line hop count, sc' represent { dr(x, y) } in foreground target average
Prospect alignment hop count, tfRepresent the parallactic angle threshold value setting, trRepresent the prospect line hop count threshold value setting, tcRepresent setting
Prospect alignment hop count threshold value.
Described step 2. middle qnAcquisition process be:
2. -1, assume that the total width number belonging to the stereo-picture of n scene mode in stereoscopic image data storehouse is m', its
In, m' >=1;
2. -2, make qnInitial value be 0.1, make qnInitial value be 0;
2. -3, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
The global disparity angle of the right viewpoint parallax gray level image of every width stereo-picture of pattern, the weight under n scene mode is taken
It is worth for qnWhen stereoscopic image data storehouse in belong to n scene mode m' width stereo-picture right viewpoint parallax gray level image
Global disparity angle be designated as dm',a(qn), dm',a(qn)=qn×|fm',a|+(1-qn)×|bm',a|, wherein, 1≤m'≤m', fm',a
And bm',aThe corresponding right viewpoint parallax ash representing the m' width stereo-picture belonging to n scene mode in stereoscopic image data storehouse
The parallactic angle of foreground target in degree image and the parallactic angle of background region, symbol " | | " it is the symbol that takes absolute value;
2. -4, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
Every width stereo-picture of pattern eliminates the comfort level objective evaluation value of parallax linear effect, by the power under n scene mode
Refetching value is qnWhen stereoscopic image data storehouse in belong to the m' width stereo-picture of n scene mode and eliminate the linear shadow of parallax
The comfort level objective evaluation value rung is designated as vch(dm',a(qn),wa), vch(dm',a(qn),wa)=4.2028+0.1912 × ln
(wa)-0.0208×dm',a(qn)×ln(wa);Then calculating the weight value under n scene mode is qnWhen stereo-picture number
According to the comfort level objective evaluation value of the parallax linear effect of the every width stereo-picture belonging to n scene mode in storehouse, by n
Weight value under scene mode is qnWhen stereoscopic image data storehouse in belong to the m' width stereo-picture of n scene mode
The comfort level objective evaluation value of parallax linear effect is designated as errm', errm'=vch(dm',a(qn),wa)-mosm', wherein, mosm'
Represent that the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to the m' width of n scene mode
The comfort level subjective assessment value of stereo-picture;Adopt least square method again, be q to the weight value under n scene modenWhen
The overall situation of the respective right viewpoint parallax gray level image of all stereo-pictures of n scene mode is belonged in stereoscopic image data storehouse
The comfort level objective evaluation value of parallactic angle and respective parallax linear effect carries out linear fit, obtains under n scene mode
Weight value be qnWhen stereoscopic image data storehouse in belong to a matching corresponding to all stereo-pictures of n scene mode
Linear equation: err=u (qn)+v(qn)×da(qn), wherein, err is the linear effect for representing parallax to objective comfort level
Dependent variable, u (qn) and v (qn) it is err=u (qn)+v(qn)×da(qn) in constant, da(qn) it is for representing n
Weight value under scene mode is qnWhen stereoscopic image data storehouse in belong to the arbitrary width stereo-picture of n scene mode
The independent variable at the global disparity angle of right viewpoint parallax gray level image;
2. it is -5, q according to the weight value under n scene modenWhen stereoscopic image data storehouse in belong to n scene
One fitting a straight line equation corresponding to all stereo-pictures of pattern, obtaining the weight value under n scene mode is qnWhen
The comfort level objective evaluation equation of all stereo-pictures of n scene mode: smmo is belonged in stereoscopic image data storehousen(qn)
=vch(da(qn),wa)-err=(4.2028-u (qn))-v(qn)×da(qn)+0.1912×ln(wa)-0.0208×da(qn)
×ln(wa), wherein, smmon(qn) it is the comfort level objective evaluation value being superimposed parallax linear effect and parallax non-linear effects,
vch(da(qn),wa) represent that the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
Arbitrary width stereo-picture of pattern eliminates the comfort level objective evaluation value of parallax linear effect;Then measure n scene mould
Weight value under formula is qnWhen stereoscopic image data storehouse in belong to n scene mode all stereo-pictures comfort level visitor
See the fitting degree of evaluation of estimate and corresponding comfort level subjective assessment value;
2. -6, make qn=qn+ 0.1, it is then back to step and 2. -3 continue executing with, until qnTerminate during equal to 1.1, obtain n-th
Plant when the weight value under scene mode is respectively 0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0 and intend accordingly
Then corresponding for best fit degree weight value is assigned to q by conjunction degreen, wherein, qn=qnIn+0.1 "=" it is assignment
Symbol;
Described step 2. middle u (qn) value be best fit degree corresponding fitting a straight line equation in a term system
Number;Described step 2. middle v (qn) value be best fit degree corresponding fitting a straight line equation in constant.
Described step is 2. in the right viewpoint parallax gray level image of every width stereo-picture in neutral body image data base
The acquisition modes of foreground target and background region and described step 3. in stereo-picture to be evaluated right viewpoint parallax gray scale
Foreground target in image is identical with the acquisition modes of background region, by the every width stereo-picture in stereoscopic image data storehouse and treating
Evaluate stereo-picture all as pending stereo-picture, then in the right viewpoint parallax gray level image of pending stereo-picture before
The acquisition process of scape target and background region is: obtains the intensity histogram of the right viewpoint parallax gray level image of pending stereo-picture
Figure;Then maximum variance between clusters are adopted, and the intensity histogram of the right viewpoint parallax gray level image according to pending stereo-picture
Figure, obtains the intensity slicing threshold value of the right viewpoint parallax gray level image of pending stereo-picture, is designated as pseg;Again will be pending vertical
In the right viewpoint parallax gray level image of body image, pixel value is more than or equal to psegPixel be defined as foreground pixel point, and will
In the right viewpoint parallax gray level image of pending stereo-picture, pixel value is less than psegPixel be defined as background pixel;?
It is made up of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture afterwards all foreground pixel points, after all
Scene vegetarian refreshments constitutes the background region in the right viewpoint parallax gray level image of pending stereo-picture.
Intensity slicing threshold value p of the right viewpoint parallax gray level image of described pending stereo-picturesegAcquisition process
For:
X1, set up intensity slicing threshold value find object function, be designated as t, t=wf×(μ-μf)2+wb×(μ-μb)2, wherein,
wfRepresent that the total number of the foreground pixel point in the right viewpoint parallax gray level image of pending stereo-picture accounts for pending stereogram
The ratio of the total number of pixel in the right viewpoint parallax gray level image of picture,μfRepresent pending vertical
The average of the pixel value of all foreground pixel points in the right viewpoint parallax gray level image of body image,wbRepresent background pixel in the right viewpoint parallax gray level image of pending stereo-picture
Total number accounts for the ratio of the total number of pixel in the right viewpoint parallax gray level image of pending stereo-picture,μbRepresent all background pixels in the right viewpoint parallax gray level image of pending stereo-picture
The average of pixel value,μ represents in the right viewpoint parallax gray level image of pending stereo-picture
The average of the pixel value of all pixels point, μ=wf×μf+wb×μb, histdG () represents that the right viewpoint of pending stereo-picture regards
Grey level histogram { the hist of difference gray level imaged(g) } in pixel value be g pixel number, 0≤gmin≤g≤gmax≤
255, gminRepresent the grey level histogram { hist of the right viewpoint parallax gray level image of pending stereo-pictured(g) } in pixel
The minimum pixel value that number is more than 0, gmaxRepresent the grey level histogram of the right viewpoint parallax gray level image of pending stereo-picture
{histd(g) } in pixel number be more than 0 max pixel value;
X2, in interval [gmin,gmax] interior traversal t, the value of t when making t maximum is defined as pending stereo-picture
Intensity slicing threshold value p of right viewpoint parallax gray level imageseg.
The described step 2. middle mode determining the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse
Identical with the mode of the scene mode belonging to described step 3. middle determination stereo-picture to be evaluated, by stereoscopic image data storehouse
In every width stereo-picture and stereo-picture to be evaluated all as pending stereo-picture it is determined that pending stereo-picture institute
The detailed process of the scene mode belonging to is:
Z1, for the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, if fa *> 1 ° it is determined that
Foreground target convex in screen and be in non-comfort zone;If 0 °≤fa *< 1 ° it is determined that foreground target convex in screen and be in comfortable
Area;If -1 ° of < fa *< 0 ° it is determined that foreground target recessed in screen and be in comfort zone;If fa *< -1 ° it is determined that foreground target is recessed
In screen and be in non-comfort zone;
For the background region in the right viewpoint parallax gray level image of pending stereo-picture, if ba *> 1 ° it is determined that background
Region convex in screen and be in non-comfort zone;If 0 °≤ba *< 1 ° it is determined that background region convex in screen and be in comfort zone;If-
1 ° of < ba *< 0 ° it is determined that background region recessed in screen and be in comfort zone;If ba *< -1 ° it is determined that background region recessed in screen
And it is in non-comfort zone;
Wherein, fa *Represent the parallactic angle of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, fa *
=f-k, ba *Represent the parallactic angle of the background region in the right viewpoint parallax gray level image of pending stereo-picture, ba *=b-k, f
Represent that human eye binocular watches the convergent angle of foreground target,After b represents the viewing of human eye binocular
The convergent angle of scene area,K represents the adjustment angle of human eye binocular,Arctan () is tan of negating, and p represents the interpupillary distance of human eye binocular, and l represents the width of display
Degree, n represents the horizontal resolution of display, and h represents human eye to the distance of display, and f represents that the right side of pending stereo-picture regards
Point parallax gray level image in foreground target mean parallax amplitude, f in units of pixel, B represents and treats
Process stereo-picture right viewpoint parallax gray level image in background region mean parallax amplitude, b in units of pixel, dr *(x,y)
Represent that in the right viewpoint parallax gray level image of pending stereo-picture, coordinate position is the pixel value of the pixel of (x, y);
Z2, concavo-convex with respect to screen according to the foreground target in the right viewpoint parallax gray level image of pending stereo-picture
Property and whether be in comfort zone, whether background region with respect to the concavity and convexity of screen and is in comfort zone, determines pending
Scene mode belonging to stereo-picture, if particularly as follows: foreground target convex in screen and be in non-comfort zone, background region recessed in screen
Curtain and be in non-comfort zone it is determined that pending stereo-picture belongs to the 1st kind of scene mode;If foreground target convex in screen and place
In non-comfort zone, background region convex in screen and be in non-comfort zone it is determined that pending stereo-picture belongs to the 2nd kind of scene mould
Formula;If foreground target recessed in screen and be in non-comfort zone, background region recessed in screen and be in non-comfort zone it is determined that waiting to locate
Reason stereo-picture belongs to the 3rd kind of scene mode;If foreground target convex in screen and be in non-comfort zone, background region recessed in screen
And it is in comfort zone it is determined that pending stereo-picture belongs to the 4th kind of scene mode;If foreground target convex in screen and be in non-
Comfort zone, background region convex in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 5th kind of scene mode;If
Foreground target convex in screen and be in comfort zone, background region recessed in screen and be in non-comfort zone it is determined that pending solid
Image belongs to the 6th kind of scene mode;If foreground target recessed in screen and be in comfort zone, background region recessed in screen and be in non-
Comfort zone is it is determined that pending stereo-picture belongs to the 7th kind of scene mode;If foreground target convex in screen and be in comfort zone,
Background region recessed in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 8th kind of scene mode;If foreground target
Convex in screen and be in comfort zone, background region convex in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 9th
Plant scene mode;If foreground target recessed in screen and be in comfort zone, background region recessed in screen and be in comfort zone it is determined that
Pending stereo-picture belongs to the 10th kind of scene mode.
Described step 3. in { dr(x, y) } in foreground target width angleWherein,
Arctan () is to negate tan, and h represents human eye to the distance of display, wfrontRepresent { dr(x, y) } in foreground target
Mean breadth, wfrontAcquisition process be:
A1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) },
The pixel value of the pixel that coordinate position in { bi (x, y) } is (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position
When belonging to foreground target for the pixel of (x, y), make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel
When point belongs to background region, make bi (x, y)=0;
A2, line scans are entered to { bi (x, y) }, for the row row in { bi (x, y) }, from the 1st picture of row row
Vegetarian refreshments starts to scan to the right, when scanning the pixel that the 1st pixel value is 1, using this pixel as the 1st section of prospect width
The original position of line segment, and the row coordinate of this pixel is designated as x1, continue scanning to the right until scanning the picture that pixel value is 0
The 1st section of prospect width line segment is obtained, the row coordinate of the pixel that this pixel value is 0 is designated as x till vegetarian refreshments2, by the 1st section of prospect
The length of width line segment is designated as wl1, wl1=x2-x1;Continue scan to the right, with obtain the 1st section of prospect width line segment identical
Mode, obtains all prospect width line segments in row row;Wherein, the initial value of row is 1,1≤row≤h, 1≤x1<x2≤
w;
A3, length in { bi (x, y) } is less than all prospect width line segments of 0.002w and length is more than the institute of 0.995w
Prospect width line segment is had to remove;Then by length, order from small to large is ranked up to remaining all prospect width line segments,
The prospect width line segment taking middle 80% constitutes prospect width line segment aggregate, is designated as { wl'n, wherein, wl'nRepresent { wl'nIn
N-th section of prospect width line segment length, 1≤n≤n', n' represent { wl'nIn total hop count of prospect width line segment of comprising;
A4, calculating { wl'nIn the length of all prospect width line segments mean value, be designated as wfront',Then by wfront' as { dr(x, y) } in foreground target mean breadth wfrontEven, wfront=
wfront'.
Described step 4. in { dr(x, y) } in foreground target average prospect line hop count
{dr(x, y) } in foreground target average prospect alignment hop countWherein, 1≤row≤h, 1≤col
≤ w, rlqrowRepresent { dr(x, y) } in row row in prospect line section quantity, clqcolRepresent { dr(x, y) } in col row in prospect alignment section quantity,
Wherein, rlqrowAnd clqcolAcquisition process be:
B1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) },
The pixel value of the pixel that coordinate position in { bi (x, y) } is (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position
When belonging to foreground target for the pixel of (x, y), make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel
When point belongs to background region, make bi (x, y)=0;
B2,2 dilation operations that { bi (x, y) } is carried out successively with mathematical morphology, 4 erosion operations, 2 expansion fortune
Calculate, obtain { bi'(x, y) }, wherein, bi'(x, y) represent that in { bi'(x, y) }, coordinate position is the pixel of the pixel of (x, y)
Value;
The quantity of effective prospect line section in often going in b3, statistics { bi'(x, y) }, by the in { bi'(x, y) }
In row row, the quantity of effective prospect line section is designated as rlqrow', rlqrow' acquisition process be: b3-1, make rlqrow' just
Initial value is 0;B3-2, the 1st pixel of row row from { bi'(x, y) } start to scan to the right, are scanning the 1st picture
During the pixel for 1 for the element value, using this pixel as the original position of the 1st section of prospect line section, and the row of this pixel are sat
It is labeled as x1', continue scanning to the right and obtain the 1st section of prospect line section till scanning the pixel that pixel value is 0, should
Pixel value is that the row coordinate of 0 pixel is designated as x2', if length x of the 1st section of prospect line section2'-x1' it is more than 0.005w, then really
Fixed 1st section of prospect line section is effective prospect line section, and makes rlqrow'=rlqrow'+1, if the 1st section of prospect line section
Length x2'-x1' it is less than or equal to 0.005w it is determined that the 1st section of prospect line section is invalid prospect line section;Continue to sweep to the right
Retouch, with obtain the 1st section of prospect line section determine whether the 1st section of prospect line section is effective prospect line section identical side
Formula, obtains all effective prospect line section in row row, and counts and obtain effective prospect line section in row row
Quantity rlqrow';Wherein, the initial value of row is 1,1≤row≤h, 1≤x1'<x2'≤w, rlqrow'=rlqrowIn '+1
"=" is assignment;
Equally, in often going in statistics { bi'(x, y) } effective prospect alignment section quantity, by { bi'(x, y) }
In col row, the quantity of effective prospect alignment section is designated as clqcol', clqcol' acquisition process be: b3-1), make clqcol'
Initial value is 0;B3-2), the 1st pixel of the col row from { bi'(x, y) } starts to scan downwards, is scanning the 1st
When individual pixel value is 1 pixel, using this pixel as the 1st section of prospect alignment section original position, and by this pixel
Row coordinate is designated as y1', continue scanning downwards and obtain the 1st section of prospect alignment section till scanning the pixel that pixel value is 0,
The row coordinate of the pixel that this pixel value is 0 is designated as y2', if length y of the 1st section of prospect alignment section2'-y1' it is more than 0.005h,
Then determine that the 1st section of prospect alignment section is effective prospect alignment section, and make clqcol'=clqcol'+1, if the 1st section of prostatitis alignment
Length y of section2'-y1' it is less than or equal to 0.005h it is determined that the 1st section of prospect alignment section is invalid prospect alignment section;Continue to
Lower scanning, with obtain the 1st section of prospect alignment section determine whether the 1st section of prospect alignment section is that effective prospect alignment section is identical
Mode, obtain all effective prospect alignment section in col row, and count and obtain in col row effectively prospect alignment
Quantity clq of sectioncol';Wherein, the initial value of col is 1,1≤col≤w, 1≤y1'<y2'≤h, clqcol'=clqcolIn '+1
"=" it is assignment;
B4, make rlqrow=rlqrow', make clqcol=clqcol'.
Described step 4. in take p=1.6, tf=2.0, tr=2, tc=1.5.
Compared with prior art, it is an advantage of the current invention that: nature according to three-dimensional display for the inventive method
In scene, it is determined that 10 kinds of scene modes of stereo-picture, this can be abundant for foreground target and the depth location residing for background region
Embody the non-linear relation of the depth factor of object and visual comfort in stereo-picture, thus based on 10 kinds of scene modes respectively
The visual comfort evaluation model set up can more accurately reflect the sense of the visual comfort to stereo-picture for the human visual system
Know such that it is able to effectively improve the uniformity between objective evaluation result and human eye subjective perception.
Brief description
Fig. 1 totally realizes block diagram for the inventive method;
Fig. 2 is the geometric representation of f, b, k, p, h, f, b;
Fig. 3 is w 'aGeometric representation;
Fig. 4 a is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.1
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 b is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.2
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 c is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.3
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 d is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.4
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 e is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.5
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 f is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.6
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 g is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.7
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 h is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.8
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 i is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 0.9
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 4 j is the weight value under the 4th kind of scene mode is to belong to the 4th kind of scene mould in stereoscopic image data storehouse when 1.0
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 5 is to belong to the 4th in stereoscopic image data storehouse during weight value respectively 0.1~1.0 under the 4th kind of scene mode
(fitting degree is weighed with mae, and weight value is for mae when 0.7 to plant fitting degree corresponding to all stereo-pictures of scene mode
Little, that is, best fit degree corresponding weight value is 0.7);
Fig. 6 a is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.1
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 b is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.2
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 c is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.3
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 d is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.4
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 e is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.5
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 f is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.6
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 g is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.7
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 h is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.8
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 i is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 0.9
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 6 j is the weight value under the 5th kind of scene mode is to belong to the 5th kind of scene mould in stereoscopic image data storehouse when 1.0
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 7 is to belong to the 5th in stereoscopic image data storehouse during weight value respectively 0.1~1.0 under the 5th kind of scene mode
(fitting degree is weighed with mae, and weight value is for mae when 0.6 to plant fitting degree corresponding to all stereo-pictures of scene mode
Little, that is, best fit degree corresponding weight value is 0.6);
Fig. 8 a is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.1
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 b is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.2
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 c is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.3
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 d is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.4
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 e is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.5
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 f is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.6
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 g is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.7
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 h is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.8
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 i is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 0.9
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 8 j is the weight value under the 8th kind of scene mode is to belong to the 8th kind of scene mould in stereoscopic image data storehouse when 1.0
Linear Fit Chart corresponding to all stereo-pictures of formula, abscissa is global disparity angle, and ordinate is the visitor of parallax linear effect
See comfortable angle value;
Fig. 9 is to belong to the 8th in stereoscopic image data storehouse during weight value respectively 0.1~1.0 under the 8th kind of scene mode
(fitting degree is weighed with mae, and weight value is for mae when 0.6 to plant fitting degree corresponding to all stereo-pictures of scene mode
Little, that is, best fit degree corresponding weight value is 0.6);
Figure 10 a is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.1
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 b is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.2
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 c is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.3
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 d is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.4
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 e is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.5
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 f is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.6
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 g is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.7
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 h is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.8
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 i is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 0.9
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 10 j is the weight value under the 9th kind of scene mode is to belong to the 9th kind of scene in stereoscopic image data storehouse when 1.0
Linear Fit Chart corresponding to all stereo-pictures of pattern, abscissa is global disparity angle, and ordinate is parallax linear effect
Objective comfortable angle value;
Figure 11 is to belong to the in stereoscopic image data storehouse when weight value under the 9th kind of scene mode is respectively 0.1~1.0
Fitting degree corresponding to all stereo-pictures of 9 kinds of scene modes (weighed with mae, and weight value is mae when 0.5 by fitting degree
Minimum, that is, best fit degree corresponding weight value is 0.5).
Specific embodiment
Below in conjunction with accompanying drawing embodiment, the present invention is described in further detail.
A kind of stereo image vision comfort level evaluation method based on scene mode classification proposed by the present invention, it is totally real
Existing block diagram is as shown in figure 1, it comprises the following steps:
1. 10 kinds of scene modes of the natural scene shown by three-dimensional display are determined, the 1st kind of scene mode is stereogram
Foreground target in the right viewpoint parallax gray level image of picture convex in screen and be in non-comfort zone, background region recessed in screen and place
In non-comfort zone;2nd kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and
Be in non-comfort zone, background region convex in screen and be in non-comfort zone;3rd kind of scene mode is that the right viewpoint of stereo-picture regards
Foreground target in difference gray level image recessed in screen and be in non-comfort zone, background region recessed in screen and be in non-comfort zone;
4th kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in non-comfortable
Area, background region recessed in screen and be in comfort zone;5th kind of scene mode is in the right viewpoint parallax gray level image of stereo-picture
Foreground target convex in screen and be in non-comfort zone, background region convex in screen and be in comfort zone;6th kind of scene mode be
Foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in comfort zone, background region recessed in screen
And it is in non-comfort zone;7th kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture recessed in screen
Curtain and be in comfort zone, background region recessed in screen and be in non-comfort zone;8th kind of scene mode is the right viewpoint of stereo-picture
Foreground target in parallax gray level image convex in screen and be in comfort zone, background region recessed in screen and be in comfort zone;9th
Kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in comfort zone, background
Region convex in screen and be in comfort zone;10th kind of scene mode is the prospect in the right viewpoint parallax gray level image of stereo-picture
Target recessed in screen and be in comfort zone, background region recessed in screen and be in comfort zone.
2. every width stereo-picture is selected to have a stereoscopic image data storehouse of the mean subjective scoring average of visual comfort;
Then the parallax according to the foreground target in the right viewpoint parallax gray level image of the every width stereo-picture in stereoscopic image data storehouse
Angle and the parallactic angle of background region, determine the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse;Further according to
Before belonging in stereoscopic image data storehouse in the respective right viewpoint parallax gray level image of all stereo-pictures of every kind of scene mode
The parallactic angle of scape target and the parallactic angle of background region, set up the visual comfort evaluation model under every kind of scene mode, by n-th
The visual comfort evaluation model planted under scene mode is described as follows: smmon=(4.2028-u (qn))-v(qn)×da(qn)+
0.1912×ln(wa)-0.0208×da(qn)×ln(wa), wherein, 1≤n≤10, smmonRepresent under n scene mode
The output of visual comfort evaluation model, u (qn) and v (qn) it is constant, u (qn) and v (qn) value by the u in step 2. -4
(qn) and v (qn) determine, da(qn) represent that a width to be entered belongs to the right side of the stereo-picture to be evaluated of n scene mode
The global disparity angle of viewpoint parallax gray level image, da(qn)=qn×|fa|+(1-qn)×|ba|, qnRepresent n scene mode
Under weight, faThe width representing to be entered belongs to the right viewpoint parallax ash of the stereo-picture to be evaluated of n scene mode
The parallactic angle of the foreground target in degree image, baThe width representing to be entered belongs to the solid to be evaluated of n scene mode
The parallactic angle of the background region in the right viewpoint parallax gray level image of image, waThe width representing to be entered belongs to n scene
The width angle of the foreground target in the right viewpoint parallax gray level image of the stereo-picture to be evaluated of pattern, symbol " | | " for taking absolutely
To value symbol.
In this particular embodiment, step 2. middle qnAcquisition process be:
2. -1, assume that the total width number belonging to the stereo-picture of n scene mode in stereoscopic image data storehouse is m', its
In, m' >=1.
2. -2, make qnInitial value be 0.1, make qnInitial value be 0.
2. -3, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
The global disparity angle of the right viewpoint parallax gray level image of every width stereo-picture of pattern, the weight under n scene mode is taken
It is worth for qnWhen stereoscopic image data storehouse in belong to n scene mode m' width stereo-picture right viewpoint parallax gray level image
Global disparity angle be designated as dm',a(qn), dm',a(qn)=qn×|fm',a|+(1-qn)×|bm',a|, wherein, 1≤m'≤m', fm',a
And bm',aThe corresponding right viewpoint parallax ash representing the m' width stereo-picture belonging to n scene mode in stereoscopic image data storehouse
The parallactic angle of foreground target in degree image and the parallactic angle of background region, symbol " | | " it is the symbol that takes absolute value.
2. -4, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
Every width stereo-picture of pattern eliminates the comfort level objective evaluation value of parallax linear effect, by the power under n scene mode
Refetching value is qnWhen stereoscopic image data storehouse in belong to the m' width stereo-picture of n scene mode and eliminate the linear shadow of parallax
The comfort level objective evaluation value rung is designated as vch(dm',a(qn),wa), vch(dm',a(qn),wa)=4.2028+0.1912 × ln
(wa)-0.0208×dm',a(qn)×ln(wa);Then calculating the weight value under n scene mode is qnWhen stereo-picture number
According to the comfort level objective evaluation value of the parallax linear effect of the every width stereo-picture belonging to n scene mode in storehouse, by n
Weight value under scene mode is qnWhen stereoscopic image data storehouse in belong to the m' width stereo-picture of n scene mode
The comfort level objective evaluation value of parallax linear effect is designated as errm', errm'=vch(dm',a(qn),wa)-mosm', wherein, mosm'
Represent that the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to the m' width of n scene mode
The comfort level subjective assessment value of stereo-picture, mosm'Value known to;Adopt existing least square method again, to n scene mould
Weight value under formula is qnWhen stereoscopic image data storehouse in belong to the respective right side of all stereo-pictures of n scene mode and regard
The comfort level objective evaluation value of the global disparity angle of point parallax gray level image and respective parallax linear effect carries out linear fit,
Obtaining the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to all vertical of n scene mode
The corresponding fitting a straight line equation of body image: err=u (qn)+v(qn)×da(qn), wherein, err is for representing parallax pair
The dependent variable of the linear effect of objective comfort level, u (qn) and v (qn) it is err=u (qn)+v(qn)×da(qn) in constant, u
(qn) and v (qn) value be matching result out, da(qn) it is for representing that the weight value under n scene mode is qnWhen
The overall situation belonging to the right viewpoint parallax gray level image of arbitrary width stereo-picture of n scene mode in stereoscopic image data storehouse regards
The independent variable of declinate.
2. it is -5, q according to the weight value under n scene modenWhen stereoscopic image data storehouse in belong to n scene
One fitting a straight line equation corresponding to all stereo-pictures of pattern, obtaining the weight value under n scene mode is qnWhen
The comfort level objective evaluation equation of all stereo-pictures of n scene mode: smmo is belonged in stereoscopic image data storehousen(qn)
=vch(da(qn),wa)-err=(4.2028-u (qn))-v(qn)×da(qn)+0.1912×ln(wa)-0.0208×da(qn)
×ln(wa), wherein, smmon(qn) it is the comfort level objective evaluation value being superimposed parallax linear effect and parallax non-linear effects,
vch(da(qn),wa) represent that the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
Arbitrary width stereo-picture of pattern eliminates the comfort level objective evaluation value of parallax linear effect;Then adopt existing averagely exhausted
Value error (mae) is measured with the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene
The fitting degree of the comfort level objective evaluation value of all stereo-pictures of pattern and corresponding comfort level subjective assessment value.
2. -6, make qn=qn+ 0.1, it is then back to step and 2. -3 continue executing with, until qnTerminate during equal to 1.1, obtain n-th
Plant when the weight value under scene mode is respectively 0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0 and intend accordingly
Then corresponding for best fit degree weight value is assigned to q by conjunction degreen, wherein, qn=qnIn+0.1 "=" it is assignment
Symbol.
In this particular embodiment, step 2. middle u (qn) value be best fit degree corresponding fitting a straight line equation
In Monomial coefficient, v (qn) value be best fit degree corresponding fitting a straight line equation in constant.
Here, the stereoscopic image data being provided with Korea Advanced Institute of Science and Technology image and video system laboratory (ivy lab)
As a example storehouse, this stereoscopic image data storehouse comprises 120 width stereo-pictures and corresponding right viewpoint parallax gray level image, this stereogram
As database contains the indoor and outdoors stereo-picture of various scene depths, and give the euphorosia of every width stereo-picture
The mean subjective scoring average of degree;Calculate the right viewpoint parallax gray level image of the every width stereo-picture in this stereoscopic image data storehouse
In the parallactic angle of foreground target and background region parallactic angle, obtain foreground target and background region each with respect to screen
Concavity and convexity and the characteristic whether being located at euphorosia area, it is possible to find the stereo-picture in this stereoscopic image data storehouse is only limitted to
4th kind, the 5th kind, the 8th kind and the 9th kind of scene mode, therefore build taking the 4th kind, the 5th kind, the 8th kind and the 9th kind of scene mode as a example
Visual comfort evaluation model under vertical 4th kind, the 5th kind, the 8th kind and the 9th kind of scene mode.
Stereo-picture when the weight value that Fig. 4 a to Fig. 4 j correspondence gives under the 4th kind of scene mode is respectively 0.1~1.0
Linear Fit Chart corresponding to all stereo-pictures of the 4th kind of scene mode is belonged to, Fig. 5 gives the 4th kind of scene mould in database
All stereo-pictures of the 4th kind of scene mode are belonged in stereoscopic image data storehouse when weight value under formula is respectively 0.1~1.0
Corresponding fitting degree (weighed with mae, and weight value is that when 0.7, mae is minimum, and that is, best fit degree is corresponding by fitting degree
Weight value is 0.7);When the weight value that Fig. 6 a to Fig. 6 j correspondence gives under the 5th kind of scene mode is respectively 0.1~1.0
Linear Fit Chart corresponding to all stereo-pictures of the 5th kind of scene mode is belonged to, Fig. 7 gives the 5th in stereoscopic image data storehouse
Plant and in stereoscopic image data storehouse when the weight value under scene mode is respectively 0.1~1.0, belong to all of the 5th kind of scene mode
Fitting degree corresponding to stereo-picture (weighed with mae, and weight value is that when 0.6, mae is minimum, i.e. best fit journey by fitting degree
Spending corresponding weight value is 0.6);The weight value that Fig. 8 a to Fig. 8 j correspondence gives under the 8th kind of scene mode is respectively 0.1
Linear Fit Chart corresponding to all stereo-pictures of the 8th kind of scene mode is belonged to, Fig. 9 gives in stereoscopic image data storehouse when~1.0
Go out and belonged to the 8th kind of scene mould in stereoscopic image data storehouse when the weight value under the 8th kind of scene mode is respectively 0.1~1.0
Fitting degree corresponding to all stereo-pictures of formula (weighed with mae, and weight value is that when 0.6, mae is minimum, that is, by fitting degree
The corresponding weight value of good fitting degree is 0.6);The weight that Figure 10 a to Figure 10 j correspondence gives under the 9th kind of scene mode takes
Linear Quasi corresponding to all stereo-pictures of the 9th kind of scene mode is belonged in stereoscopic image data storehouse when value is respectively 0.1~1.0
Close figure, belong in stereoscopic image data storehouse when the weight value that Figure 11 gives under the 9th kind of scene mode is respectively 0.1~1.0
Fitting degree corresponding to all stereo-pictures of the 9th kind of scene mode (weighed with mae, when weight value is 0.5 by fitting degree
Mae is minimum, and that is, best fit degree corresponding weight value is 0.5).
It is respectively as follows: q according to calculating the weight under the 4th kind, the 5th kind, the 8th kind and the 9th kind of scene mode above4=0.7, q5
=0.6, q8=0.6, q9=0.5, the visual comfort evaluation model under the 4th kind, the 5th kind, the 8th kind and the 9th kind of scene mode divides
It is not:
smmo4=4.3938-0.6652 × da(q4)+0.1912×ln(wa)-0.0208×da(q4)×ln(wa)
smmo5=4.2326-0.7120 × da(q5)+0.1912×ln(wa)-0.0208×da(q5)×ln(wa)
smmo8=4.5232-0.8918 × da(q8)+0.1912×ln(wa)-0.0208×da(q8)×ln(wa)
smmo9=5.4614-2.6905 × da(q9)+0.1912×ln(wa)-0.0208×da(q9)×ln(wa)
3. the left view point coloured image of stereo-picture to be evaluated is designated as { il(x, y) }, by stereo-picture to be evaluated
Right viewpoint coloured image be designated as { ir(x, y) }, the right viewpoint parallax gray level image of stereo-picture to be evaluated is designated as { dr(x,
Y) }, wherein, 1≤x≤w, 1≤y≤h, w represent the width of stereo-picture to be evaluated, its with stereoscopic image data storehouse in every
The width of width stereo-picture is consistent, and h represents the height of stereo-picture to be evaluated, and it is stood with the every width in stereoscopic image data storehouse
Highly consistent, the i of body imagel(x, y) represents { il(x, y) } in coordinate position be (x, y) pixel pixel value, ir(x,y)
Represent { ir(x, y) } in coordinate position be (x, y) pixel pixel value, dr(x, y) represents { dr(x, y) } in coordinate position
Pixel value for the pixel of (x, y);Then adopt and step 2. middle identical mode, determine belonging to stereo-picture to be evaluated
Scene mode;Further according to the scene mode belonging to stereo-picture to be evaluated, choose the visual comfort under this scene mode
Evaluation model;Then according to the visual comfort evaluation model under this scene mode and { dr(x, y) } in the regarding of foreground target
Declinate, { dr(x, y) } in the parallactic angle of background region and { dr(x, y) } in foreground target width angle, calculate to be evaluated
The visual comfort evaluation and foreca value of stereo-picture, is designated as smmo it is assumed that stereo-picture to be evaluated belongs to n scene mould
Formula, then smmo=(4.2028-u (qn))-v(qn)×da'(qn)+0.1912×ln(wa')-0.0208×da'(qn)×ln
(wa'), wherein, da'(qn) represent { dr(x, y) } global disparity angle, da'(qn)=qn×|fa'|+(1-qn)×|ba' |, fa'
Represent { dr(x, y) } in foreground target parallactic angle, ba' represent { dr(x, y) } in background region parallactic angle, wa' represent
{dr(x, y) } in foreground target width angle.
In this particular embodiment, step 3. in { dr(x, y) } in foreground target width angleWherein, arctan () is to negate tan, and h represents human eye to the distance of display, wfront
Represent { dr(x, y) } in foreground target mean breadth, wfrontAcquisition process be:
A1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) },
The pixel value of the pixel that coordinate position in { bi (x, y) } is (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position
When belonging to foreground target for the pixel of (x, y), make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel
When point belongs to background region, make bi (x, y)=0.
A2, line scans are entered to { bi (x, y) }, for the row row in { bi (x, y) }, from the 1st picture of row row
Vegetarian refreshments starts to scan to the right, when scanning the pixel that the 1st pixel value is 1, using this pixel as the 1st section of prospect width
The original position of line segment, and the row coordinate of this pixel is designated as x1, continue scanning to the right until scanning the picture that pixel value is 0
The 1st section of prospect width line segment is obtained, the row coordinate of the pixel that this pixel value is 0 is designated as x till vegetarian refreshments2, by the 1st section of prospect
The length of width line segment is designated as wl1, wl1=x2-x1;Continue scan to the right, with obtain the 1st section of prospect width line segment identical
Mode, obtains all prospect width line segments in row row;Wherein, the initial value of row is 1,1≤row≤h, 1≤x1<x2≤
w.
A3, remove the abnormal prospect width line segment of length in { bi (x, y) }, that is, remove especially short prospect width line segment and
Especially long prospect width line segment, will length be less than all prospect width line segments of 0.002w and length is big in { bi (x, y) }
All prospect width line segments in 0.995w remove;Then press length order from small to large to remaining all prospect wide lines
Section is ranked up, and takes middle 80% prospect width line segment to constitute prospect width line segment aggregate, is designated as { wl'n, wherein, wl'nTable
Show { wl'nIn n-th section of prospect width line segment length, 1≤n≤n', n' represent { wl'nIn the prospect width line segment that comprises
Total hop count.
A4, calculating { wl'nIn the length of all prospect width line segments mean value, be designated as wfront',Then by wfront' as { dr(x, y) } in foreground target mean breadth wfrontEven, wfront=
wfront'.
Fig. 3 gives wa' geometric representation.
4. visual comfort evaluation and foreca value smmo of stereo-picture to be evaluated is modified, by stereo-picture to be evaluated
Revised visual comfort evaluation and foreca value be designated as smm,
Wherein, max () is to take max function, and p represents that the visual comfort under the scene mode belonging to stereo-picture to be evaluated is commented
The tortuosity attenuation coefficient of valency model, sr' represent { dr(x, y) } in foreground target average prospect line hop count, sc' represent
{dr(x, y) } in foreground target average prospect alignment hop count, tfRepresent the parallactic angle threshold value setting, trRepresent before setting
Scape line hop count threshold value, tcRepresent the prospect alignment hop count threshold value setting, take p=1.6, t in the present embodimentf=2.0, tr=
2、tc=1.5.
In this particular embodiment, step 4. in { dr(x, y) } in foreground target average prospect line hop count {dr(x, y) } in foreground target average prospect alignment hop count Its
In, sr' and sc' it is { dr(x, y) } in foreground target tortuosity parameter, 1≤row≤h, 1≤col≤w, rlqrowRepresent
{dr(x, y) } in row row in prospect line section quantity, clqcolRepresent { dr
(x, y) } in col row in prospect alignment section quantity,
Wherein, rlqrowAnd clqcolAcquisition process be:
B1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) },
The pixel value of the pixel that coordinate position in { bi (x, y) } is (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position
When belonging to foreground target for the pixel of (x, y), make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel
When point belongs to background region, make bi (x, y)=0.
B2,2 dilation operations that { bi (x, y) } is carried out successively with mathematical morphology, 4 erosion operations, 2 expansion fortune
Calculate, obtain { bi'(x, y) }, wherein, bi'(x, y) represent that in { bi'(x, y) }, coordinate position is the pixel of the pixel of (x, y)
Value.
The quantity of effective prospect line section in often going in b3, statistics { bi'(x, y) }, by the in { bi'(x, y) }
In row row, the quantity of effective prospect line section is designated as rlqrow', rlqrow' acquisition process be: b3-1, make rlqrow' just
Initial value is 0;B3-2, the 1st pixel of row row from { bi'(x, y) } start to scan to the right, are scanning the 1st picture
During the pixel for 1 for the element value, using this pixel as the original position of the 1st section of prospect line section, and the row of this pixel are sat
It is labeled as x1', continue scanning to the right and obtain the 1st section of prospect line section till scanning the pixel that pixel value is 0, should
Pixel value is that the row coordinate of 0 pixel is designated as x2', if length x of the 1st section of prospect line section2'-x1' it is more than 0.005w, then really
Fixed 1st section of prospect line section is effective prospect line section, and makes rlqrow'=rlqrow'+1, if the 1st section of prospect line section
Length x2'-x1' it is less than or equal to 0.005w it is determined that the 1st section of prospect line section is invalid prospect line section;Continue to sweep to the right
Retouch, with obtain the 1st section of prospect line section determine whether the 1st section of prospect line section is effective prospect line section identical side
Formula, obtains all effective prospect line section in row row, and counts and obtain effective prospect line section in row row
Quantity rlqrow';Wherein, the initial value of row is 1,1≤row≤h, 1≤x1'<x2'≤w, rlqrow'=rlqrowIn '+1
"=" is assignment.
Equally, in often going in statistics { bi'(x, y) } effective prospect alignment section quantity, by { bi'(x, y) }
In col row, the quantity of effective prospect alignment section is designated as clqcol', clqcol' acquisition process be: b3-1), make clqcol'
Initial value is 0;B3-2), the 1st pixel of the col row from { bi'(x, y) } starts to scan downwards, is scanning the 1st
When individual pixel value is 1 pixel, using this pixel as the 1st section of prospect alignment section original position, and by this pixel
Row coordinate is designated as y1', continue scanning downwards and obtain the 1st section of prospect alignment section till scanning the pixel that pixel value is 0,
The row coordinate of the pixel that this pixel value is 0 is designated as y2', if length y of the 1st section of prospect alignment section2'-y1' it is more than 0.005h,
Then determine that the 1st section of prospect alignment section is effective prospect alignment section, and make clqcol'=clqcol'+1, if the 1st section of prostatitis alignment
Length y of section2'-y1' it is less than or equal to 0.005h it is determined that the 1st section of prospect alignment section is invalid prospect alignment section;Continue to
Lower scanning, with obtain the 1st section of prospect alignment section determine whether the 1st section of prospect alignment section is that effective prospect alignment section is identical
Mode, obtain all effective prospect alignment section in col row, and count and obtain in col row effectively prospect alignment
Quantity clq of sectioncol';Wherein, the initial value of col is 1,1≤col≤w, 1≤y1'<y2'≤h, clqcol'=clqcolIn '+1
"=" it is assignment.
B4, make rlqrow=rlqrow', make clqcol=clqcol'.
In above-mentioned, step is 2. in the right viewpoint parallax gray level image of every width stereo-picture in neutral body image data base
Foreground target and background region acquisition modes and step 3. in stereo-picture to be evaluated right viewpoint parallax gray level image
In foreground target identical with the acquisition modes of background region, by the every width stereo-picture in stereoscopic image data storehouse and to be evaluated
Stereo-picture all as pending stereo-picture, then the prospect mesh in the right viewpoint parallax gray level image of pending stereo-picture
The acquisition process of mark and background region is: obtains the grey level histogram of the right viewpoint parallax gray level image of pending stereo-picture;
Then existing maximum variance between clusters (otsu) are adopted, and according to the right viewpoint parallax gray level image of pending stereo-picture
Grey level histogram, obtains the intensity slicing threshold value of the right viewpoint parallax gray level image of pending stereo-picture, is designated as pseg;Again will
In the right viewpoint parallax gray level image of pending stereo-picture, pixel value is more than or equal to psegPixel be defined as foreground pixel
Point, and pixel value in the right viewpoint parallax gray level image of pending stereo-picture is less than psegPixel be defined as background picture
Vegetarian refreshments;Finally it is made up of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture all foreground pixel points,
It is made up of the background region in the right viewpoint parallax gray level image of pending stereo-picture all background pixels.
Here, intensity slicing threshold value p of the right viewpoint parallax gray level image of pending stereo-picturesegAcquisition process be:
X1, set up intensity slicing threshold value find object function, be designated as t, t=wf×(μ-μf)2+wb×(μ-μb)2, wherein, wf
Represent that the total number of the foreground pixel point in the right viewpoint parallax gray level image of pending stereo-picture accounts for pending stereo-picture
The ratio of the total number of pixel in right viewpoint parallax gray level image,μfRepresent pending stereogram
The average of the pixel value of all foreground pixel points in the right viewpoint parallax gray level image of picture,wb
Represent that the total number of the background pixel in the right viewpoint parallax gray level image of pending stereo-picture accounts for pending stereo-picture
The ratio of the total number of pixel in right viewpoint parallax gray level image,μbRepresent pending stereogram
The average of the pixel value of all background pixels in the right viewpoint parallax gray level image of picture,μ
Represent the average of the pixel value of all pixels point in the right viewpoint parallax gray level image of pending stereo-picture, μ=wf×μf+
wb×μb, histdG () represents the grey level histogram { hist of the right viewpoint parallax gray level image of pending stereo-pictured(g) } in
Pixel value is the number of the pixel of g, 0≤gmin≤g≤gmax≤ 255, gminRepresent the right viewpoint parallax of pending stereo-picture
Grey level histogram { the hist of gray level imaged(g) } in pixel number be more than 0 minimum pixel value, gmaxRepresent pending vertical
Grey level histogram { the hist of the right viewpoint parallax gray level image of body imaged(g) } in pixel number be more than 0 maximum pixel
Value;Grey level histogram { the hist of the right viewpoint parallax gray level image of pending stereo-pictured(g) } in abscissa represent 0~
255 gray level, ordinate represent that the number of pixel, each pillar represent the number of the pixel of different grey-scale.
X2, in interval [gmin,gmax] interior traversal t, the value of t when making t maximum is defined as pending stereo-picture
Intensity slicing threshold value p of right viewpoint parallax gray level imageseg.
In above-mentioned, the step 2. middle side determining the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse
Formula is 3. middle with step to determine that the mode of the scene mode belonging to stereo-picture to be evaluated is identical, by stereoscopic image data storehouse
Every width stereo-picture and stereo-picture to be evaluated are all as pending stereo-picture it is determined that belonging to pending stereo-picture
The detailed process of scene mode is:
Z1, for the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, if fa *> 1 ° it is determined that
Foreground target convex in screen and be in non-comfort zone;If 0 °≤fa *< 1 ° it is determined that foreground target convex in screen and be in comfortable
Area;If -1 ° of < fa *< 0 ° it is determined that foreground target recessed in screen and be in comfort zone;If fa *< -1 ° it is determined that foreground target is recessed
In screen and be in non-comfort zone.
For the background region in the right viewpoint parallax gray level image of pending stereo-picture, if ba *> 1 ° it is determined that background
Region convex in screen and be in non-comfort zone;If 0 °≤ba *< 1 ° it is determined that background region convex in screen and be in comfort zone;If-
1°<ba *< 0 ° it is determined that background region recessed in screen and be in comfort zone;If ba *< -1 ° it is determined that background region recessed in screen
And it is in non-comfort zone.
Wherein, fa *Represent the parallactic angle of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, fa *=
F-k, ba *Represent the parallactic angle of the background region in the right viewpoint parallax gray level image of pending stereo-picture, ba *=b-k, f table
A binocular of leting others have a look at watches the convergent angle of foreground target,B represents human eye binocular viewing background
The convergent angle in region,K represents the adjustment angle of human eye binocular,
Arctan () is tan of negating, and p represents the interpupillary distance of human eye binocular, and l represents the width of display, and n represents the water of display
Divide resolution equally, h represents human eye to the distance of display, f represents in the right viewpoint parallax gray level image of pending stereo-picture
The mean parallax amplitude of foreground target, f in units of pixel, B represents and treats
Process stereo-picture right viewpoint parallax gray level image in background region mean parallax amplitude, b in units of pixel, dr *(x,y)
Represent that in the right viewpoint parallax gray level image of pending stereo-picture, coordinate position is the pixel value of the pixel of (x, y), Fig. 2 gives
Go out the geometric representation of f, b, k, p, h, f, b.
Z2, concavo-convex with respect to screen according to the foreground target in the right viewpoint parallax gray level image of pending stereo-picture
Property and whether be in comfort zone, whether background region with respect to the concavity and convexity of screen and is in comfort zone, determines pending
Scene mode belonging to stereo-picture, if particularly as follows: foreground target convex in screen and be in non-comfort zone, background region recessed in screen
Curtain and be in non-comfort zone, i.e. fa *> 1 ° and ba *< -1 ° it is determined that pending stereo-picture belongs to the 1st kind of scene mode;If front
Scape target convex in screen and be in non-comfort zone, background region convex in screen and be in non-comfort zone, i.e. fa *> 1 ° and ba *> 1 °,
Then determine that pending stereo-picture belongs to the 2nd kind of scene mode;If foreground target recessed in screen and be in non-comfort zone, rear scenic spot
Domain recessed in screen and be in non-comfort zone, i.e. fa *< -1 ° and ba *< -1 ° it is determined that pending stereo-picture belongs to the 3rd kind of scene
Pattern;If foreground target convex in screen and be in non-comfort zone, background region recessed in screen and be in comfort zone, i.e. fa *> 1 ° and-
1°≤ba *< 0 ° it is determined that pending stereo-picture belongs to the 4th kind of scene mode;If foreground target convex in screen and be in non-relax
Suitable area, background region convex in screen and be in comfort zone, i.e. fa *> 1 ° and 0 °≤ba *≤ 1 ° it is determined that pending stereo-picture belongs to
In the 5th kind of scene mode;If foreground target convex in screen and be in comfort zone, background region recessed in screen and be in non-comfortable
Area, i.e. 0 °≤fa *≤ 1 ° and ba *< -1 ° it is determined that pending stereo-picture belongs to the 6th kind of scene mode;If foreground target recessed in
Screen and be in comfort zone, background region recessed in screen and be in non-comfort zone, that is, -1 °≤fa *< 0 ° and ba *< -1 ° it is determined that
Pending stereo-picture belongs to the 7th kind of scene mode;If foreground target convex in screen and be in comfort zone, background region recessed in screen
Curtain and be in comfort zone, i.e. 0 °≤fa *≤ 1 ° and -1 °≤ba *< 0 ° it is determined that pending stereo-picture belongs to the 8th kind of scene mould
Formula;If foreground target convex in screen and be in comfort zone, background region convex in screen and be in comfort zone, i.e. 0 °≤fa *≤ 1 ° and
0°≤ba *≤ 1 ° it is determined that pending stereo-picture belongs to the 9th kind of scene mode;If foreground target recessed in screen and be in comfortable
Area, background region recessed in screen and be in comfort zone, that is, -1 °≤fa *< 0 ° and -1 °≤ba *< 0 ° it is determined that pending stereogram
As belonging to the 10th kind of scene mode.
In the present embodiment, objective parameters are commonly used as evaluation index by the use of assessment 5 of image quality evaluating method, that is,
Pearson coefficient correlation under the conditions of nonlinear regression (pearson linear correlation coefficient,
Plcc), spearman coefficient correlation (spearman rank order correlation coefficient, srocc),
Kendall coefficient correlation (kendall rank-order correlation coefficient, krocc), average absolute value
Error (mean absolute error, mae), root-mean-square error (root mean squared error, rmse).Plcc is anti-
Reflect the correlation of objective evaluation predicted value and subjective assessment value;Srocc and krocc reflection objective evaluation predicted value and subjective assessment
The monotonicity of value, uniformity;Mae and rmse reflects the accuracy of objective evaluation predicted value.Obtain 120 width using the inventive method
The respective final visual comfort evaluation and foreca value of stereo-picture, the final visual comfort to this 120 width stereo-picture
Evaluation and foreca value does five parameter logistic function nonlinear fittings, and plcc, srocc and krocc value is higher, and mae and rmse
Correlation between the evaluation result of value less explanation the inventive method and mean subjective scoring average is better.Table 1 gives this
The evaluation index value of inventive method, illustrates the result of the objective evaluation result and human eye subjective perception obtaining using the inventive method
More consistent, illustrate the validity of the inventive method well.
Table 1 evaluation index value
Evaluation index | plcc | srocc | krocc | mae | rmse |
0.919 | 0.909 | 0.748 | 0.237 | 0.318 |
Claims (8)
1. a kind of stereo image vision comfort level evaluation method based on scene mode classification is it is characterised in that include following walking
Rapid:
1. 10 kinds of scene modes of the natural scene shown by three-dimensional display are determined, the 1st kind of scene mode is stereo-picture
Foreground target in right viewpoint parallax gray level image convex in screen and be in non-comfort zone, background region recessed in screen and be in non-
Comfort zone;2nd kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in
Non- comfort zone, background region convex in screen and be in non-comfort zone;3rd kind of scene mode is the right viewpoint parallax ash of stereo-picture
Foreground target in degree image recessed in screen and be in non-comfort zone, background region recessed in screen and be in non-comfort zone;4th kind
Scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and be in non-comfort zone, background
Region recessed in screen and be in comfort zone;5th kind of scene mode is the prospect in the right viewpoint parallax gray level image of stereo-picture
Target convex in screen and be in non-comfort zone, background region convex in screen and be in comfort zone;6th kind of scene mode is stereogram
Foreground target in the right viewpoint parallax gray level image of picture convex in screen and be in comfort zone, background region recessed in screen and be in
Non- comfort zone;7th kind of scene mode be foreground target in the right viewpoint parallax gray level image of stereo-picture recessed in screen and place
In comfort zone, background region recessed in screen and be in non-comfort zone;8th kind of scene mode is the right viewpoint parallax ash of stereo-picture
Foreground target in degree image convex in screen and be in comfort zone, background region recessed in screen and be in comfort zone;9th kind of scene
Pattern be foreground target in the right viewpoint parallax gray level image of stereo-picture convex in screen and to be in comfort zone, background region convex
In screen and be in comfort zone;10th kind of scene mode is that the foreground target in the right viewpoint parallax gray level image of stereo-picture is recessed
In screen and be in comfort zone, background region recessed in screen and be in comfort zone;
Here, for the foreground target in the right viewpoint parallax gray level image of stereo-picture, if fa *> 1 ° it is determined that foreground target
Convex in screen and be in non-comfort zone;If 0 °≤fa *< 1 ° it is determined that foreground target convex in screen and be in comfort zone;If -1 ° <
fa *< 0 ° it is determined that foreground target recessed in screen and be in comfort zone;If fa *< -1 ° it is determined that foreground target recessed in screen and place
In non-comfort zone;
For the background region in the right viewpoint parallax gray level image of stereo-picture, if ba *> 1 ° it is determined that background region convex in screen
Curtain and be in non-comfort zone;If 0 °≤ba *< 1 ° it is determined that background region convex in screen and be in comfort zone;If -1 ° of <ba *< 0 °,
Then determine background region recessed in screen and be in comfort zone;If ba *< -1 ° it is determined that background region recessed in screen and be in non-relax
Suitable area;
Wherein, fa *Represent the parallactic angle of the foreground target in the right viewpoint parallax gray level image of stereo-picture, fa *=f-k, ba *Table
Show the parallactic angle of the background region in the right viewpoint parallax gray level image of stereo-picture, ba *=b-k, f represent that human eye binocular is watched
The convergent angle of foreground target,B represents that human eye binocular watches the convergent angle of background region,K represents the adjustment angle of human eye binocular,Arctan () is just to negate
Cut function, p represents the interpupillary distance of human eye binocular, l represents the width of display, n represents the horizontal resolution of display, h represents that human eye arrives
The distance of display, f represents the mean parallax amplitude of the foreground target in the right viewpoint parallax gray level image of stereo-picture, and f is with picture
Vegetarian refreshments is unit,
B represents the mean parallax amplitude of the background region in the right viewpoint parallax gray level image of stereo-picture, b in units of pixel,dr *(x, y) table
Show that in the right viewpoint parallax gray level image of stereo-picture, coordinate position is the pixel value of the pixel of (x, y);
2. every width stereo-picture is selected to have a stereoscopic image data storehouse of the mean subjective scoring average of visual comfort;Then
Parallactic angle according to the foreground target in the right viewpoint parallax gray level image of the every width stereo-picture in stereoscopic image data storehouse and
The parallactic angle of background region, determines the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse;Further according to solid
The prospect mesh in the respective right viewpoint parallax gray level image of all stereo-pictures of every kind of scene mode is belonged in image data base
Target parallactic angle and the parallactic angle of background region, set up the visual comfort evaluation model under every kind of scene mode, by n field
Visual comfort evaluation model under scape pattern is described as follows: smmon=(4.2028-u (qn))-v(qn)×da(qn)+0.1912
×ln(wa)-0.0208×da(qn)×ln(wa), wherein, 1≤n≤10, smmonRepresent that the vision under n scene mode is relaxed
The output of appropriate evaluation model, u (qn) and v (qn) it is constant, da(qn) represent that a width to be entered belongs to n scene mould
The global disparity angle of the right viewpoint parallax gray level image of the stereo-picture to be evaluated of formula, da(qn)=qn×|fa|+(1-qn)×|
ba|, qnRepresent the weight under n scene mode, faThe width representing to be entered belongs to the to be evaluated of n scene mode
The parallactic angle of the foreground target in the right viewpoint parallax gray level image of stereo-picture, baThe width representing to be entered belongs to n
The parallactic angle of the background region in the right viewpoint parallax gray level image of the stereo-picture to be evaluated of scene mode, waExpression is treated defeated
The width entering belongs to the foreground target in the right viewpoint parallax gray level image of the stereo-picture to be evaluated of n scene mode
Width angle, symbol " | | " it is the symbol that takes absolute value;
3. the right viewpoint parallax gray level image of stereo-picture to be evaluated is designated as { dr(x, y) }, wherein, 1≤x≤w, 1≤y≤
H, w represent the width of stereo-picture to be evaluated, and it is consistent with the width of the every width stereo-picture in stereoscopic image data storehouse, h
Represent the height of stereo-picture to be evaluated, its with stereoscopic image data storehouse in every width stereo-picture highly consistent, dr(x,
Y) represent { dr(x, y) } in coordinate position be (x, y) pixel pixel value;Then adopt and step 2. middle identical side
Formula, determines the scene mode belonging to stereo-picture to be evaluated;Further according to the scene mode belonging to stereo-picture to be evaluated, select
Take the visual comfort evaluation model under this scene mode;Then according to the visual comfort evaluation model under this scene mode and
{dr(x, y) } in the parallactic angle of foreground target, { dr(x, y) } in the parallactic angle of background region and { dr(x, y) } in prospect
The width angle of target, calculates the visual comfort evaluation and foreca value of stereo-picture to be evaluated, is designated as smmo it is assumed that to be evaluated
Stereo-picture belongs to n scene mode, then smmo=(4.2028-u (qn))-v(qn)×da'(qn)+0.1912×ln
(wa')-0.0208×da'(qn)×ln(wa'), wherein, da'(qn) represent { dr(x, y) } global disparity angle, da'(qn)=qn
×|fa'|+(1-qn)×|ba' |, fa' represent { dr(x, y) } in foreground target parallactic angle, ba' represent { dr(x, y) } in
The parallactic angle of background region, wa' represent { dr(x, y) } in foreground target width angle;
4. visual comfort evaluation and foreca value smmo of stereo-picture to be evaluated is modified, by repairing of stereo-picture to be evaluated
Visual comfort evaluation and foreca value after just is designated as smm,
Wherein, max () is to take max function, and p represents that the visual comfort under the scene mode belonging to stereo-picture to be evaluated is commented
The tortuosity attenuation coefficient of valency model, sr' represent { dr(x, y) } in foreground target average prospect line hop count, sc' represent
{dr(x, y) } in foreground target average prospect alignment hop count, tfRepresent the parallactic angle threshold value setting, trRepresent before setting
Scape line hop count threshold value, tcRepresent the prospect alignment hop count threshold value setting.
2. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 1, its
It is characterised by described step 2. middle qnAcquisition process be:
2. -1, assume that the total width number belonging to the stereo-picture of n scene mode in stereoscopic image data storehouse is m', wherein, m'
≥1;
2. -2, make qnInitial value be 0.1, make qnInitial value be 0;
2. -3, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene mode
The right viewpoint parallax gray level image of every width stereo-picture global disparity angle, the weight value under n scene mode is
qnWhen stereoscopic image data storehouse in belong to n scene mode the right viewpoint parallax gray level image of m' width stereo-picture complete
Office's parallactic angle is designated as dm',a(qn), dm',a(qn)=qn×|fm',a|+(1-qn)×|bm',a|, wherein, 1≤m'≤m', fm',aWith
bm',aThe corresponding right viewpoint parallax gray scale representing the m' width stereo-picture belonging to n scene mode in stereoscopic image data storehouse
The parallactic angle of the foreground target in image and the parallactic angle of background region, symbol " | | " it is the symbol that takes absolute value;
2. -4, calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene mode
Every width stereo-picture eliminate the comfort level objective evaluation value of parallax linear effect, the weight under n scene mode is taken
It is worth for qnWhen stereoscopic image data storehouse in belong to the m' width stereo-picture of n scene mode and eliminate parallax linear effect
Comfort level objective evaluation value is designated as vch(dm',a(qn),wa), vch(dm',a(qn),wa)=4.2028+0.1912 × ln (wa)-
0.0208×dm',a(qn)×ln(wa);Then calculating the weight value under n scene mode is qnWhen stereoscopic image data storehouse
In belong to n scene mode the parallax linear effect of every width stereo-picture comfort level objective evaluation value, by n scene
Weight value under pattern is qnWhen stereoscopic image data storehouse in belong to n scene mode m' width stereo-picture parallax
The comfort level objective evaluation value of linear effect is designated as errm', errm'=vch(dm'<a(qn),wa)-mosm', wherein, mosm'Represent
Weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene mode m' width three-dimensional
The comfort level subjective assessment value of image;Adopt least square method again, be q to the weight value under n scene modenShi Liti
The global disparity of the respective right viewpoint parallax gray level image of all stereo-pictures of n scene mode is belonged in image data base
The comfort level objective evaluation value of angle and respective parallax linear effect carries out linear fit, obtains the power under n scene mode
Refetching value is qnWhen stereoscopic image data storehouse in belong to a fitting a straight line corresponding to all stereo-pictures of n scene mode
Equation: err=u (qn)+v(qn)×da(qn), wherein, err be for represent parallax to the linear effect of objective comfort level because
Variable, u (qn) and v (qn) it is err=u (qn)+v(qn)×da(qn) in constant, da(qn) it is for representing n scene
Weight value under pattern is qnWhen stereoscopic image data storehouse in belong to n scene mode the right side of arbitrary width stereo-picture regard
The independent variable at the global disparity angle of point parallax gray level image;
2. it is -5, q according to the weight value under n scene modenWhen stereoscopic image data storehouse in belong to n scene mode
A fitting a straight line equation corresponding to all stereo-pictures, obtaining the weight value under n scene mode is qnShi Liti
The comfort level objective evaluation equation of all stereo-pictures of n scene mode: smmo is belonged in image data basen(qn)=vch
(da(qn),wa)-err=(4.2028-u (qn))-v(qn)×da(qn)+0.1912×ln(wa)-0.0208×da(qn)×ln
(wa), wherein, smmon(qn) it is the comfort level objective evaluation value being superimposed parallax linear effect and parallax non-linear effects, vch
(da(qn),wa) represent that the weight value under n scene mode is qnWhen stereoscopic image data storehouse in belong to n scene mould
Arbitrary width stereo-picture of formula eliminates the comfort level objective evaluation value of parallax linear effect;Then measure n scene mode
Under weight value be qnWhen stereoscopic image data storehouse in belong to n scene mode all stereo-pictures comfort level objective
The fitting degree of evaluation of estimate and corresponding comfort level subjective assessment value;
2. -6, make qn=qn+ 0.1, it is then back to step and 2. -3 continue executing with, until qnTerminate during equal to 1.1, obtain n field
Corresponding matching journey when weight value under scape pattern is respectively 0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0
Then corresponding for best fit degree weight value is assigned to q by degreen, wherein, qn=qnIn+0.1 "=" it is assignment;
Described step 2. middle u (qn) value be best fit degree corresponding fitting a straight line equation in Monomial coefficient;Institute
The step stated 2. middle v (qn) value be best fit degree corresponding fitting a straight line equation in constant.
3. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 1 and 2,
It is characterized in that described step is 2. in the right viewpoint parallax gray level image of every width stereo-picture in neutral body image data base
Foreground target and background region acquisition modes and described step 3. in stereo-picture to be evaluated right viewpoint parallax ash
Foreground target in degree image is identical with the acquisition modes of background region, by the every width stereo-picture in stereoscopic image data storehouse and
Stereo-picture to be evaluated all as pending stereo-picture, then in the right viewpoint parallax gray level image of pending stereo-picture
The acquisition process of foreground target and background region is: the gray scale obtaining the right viewpoint parallax gray level image of pending stereo-picture is straight
Fang Tu;Then maximum variance between clusters are adopted, and straight according to the gray scale of the right viewpoint parallax gray level image of pending stereo-picture
Fang Tu, obtains the intensity slicing threshold value of the right viewpoint parallax gray level image of pending stereo-picture, is designated as pseg;Again will be pending
In the right viewpoint parallax gray level image of stereo-picture, pixel value is more than or equal to psegPixel be defined as foreground pixel point, and
Pixel value in the right viewpoint parallax gray level image of pending stereo-picture is less than psegPixel be defined as background pixel;
Foreground target in the last right viewpoint parallax gray level image being made up of pending stereo-picture all foreground pixel points, by owning
Background pixel constitutes the background region in the right viewpoint parallax gray level image of pending stereo-picture.
4. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 3, its
It is characterised by intensity slicing threshold value p of the described right viewpoint parallax gray level image of pending stereo-picturesegAcquisition process
For:
X1, set up intensity slicing threshold value find object function, be designated as t, t=wf×(μ-μf)2+wb×(μ-μb)2, wherein, wfRepresent
The total number of the foreground pixel point in the right viewpoint parallax gray level image of pending stereo-picture accounts for the right side of pending stereo-picture
The ratio of the total number of pixel in viewpoint parallax gray level image,μfRepresent pending stereo-picture
Right viewpoint parallax gray level image in the pixel value of all foreground pixel points average,wbTable
Show that the total number of the background pixel in the right viewpoint parallax gray level image of pending stereo-picture accounts for pending stereo-picture
The ratio of the total number of pixel in right viewpoint parallax gray level image,μbRepresent pending stereogram
The average of the pixel value of all background pixels in the right viewpoint parallax gray level image of picture,μ
Represent the average of the pixel value of all pixels point in the right viewpoint parallax gray level image of pending stereo-picture, μ=wf×μf+
wb×μb, histdG () represents the grey level histogram { hist of the right viewpoint parallax gray level image of pending stereo-pictured(g) } in
Pixel value is the number of the pixel of g, 0≤gmin≤g≤gmax≤ 255, gminRepresent the right viewpoint parallax of pending stereo-picture
Grey level histogram { the hist of gray level imaged(g) } in pixel number be more than 0 minimum pixel value, gmaxRepresent pending vertical
Grey level histogram { the hist of the right viewpoint parallax gray level image of body imaged(g) } in pixel number be more than 0 maximum pixel
Value;
X2, in interval [gmin,gmax] interior traversal t, the right side that the value of t when making t maximum is defined as pending stereo-picture regards
Intensity slicing threshold value p of point parallax gray level imageseg.
5. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 3, its
It is characterised by the described step 2. middle mode determining the scene mode belonging to every width stereo-picture in stereoscopic image data storehouse
Identical with the mode of the scene mode belonging to described step 3. middle determination stereo-picture to be evaluated, by stereoscopic image data storehouse
In every width stereo-picture and stereo-picture to be evaluated all as pending stereo-picture it is determined that pending stereo-picture institute
The detailed process of the scene mode belonging to is:
Z1, for the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, if fa *> 1 ° it is determined that prospect
Target convex in screen and be in non-comfort zone;If 0 °≤fa *< 1 ° it is determined that foreground target convex in screen and be in comfort zone;If-
1°<fa *< 0 ° it is determined that foreground target recessed in screen and be in comfort zone;If fa *< -1 ° it is determined that foreground target recessed in screen
And it is in non-comfort zone;
For the background region in the right viewpoint parallax gray level image of pending stereo-picture, if ba *> 1 ° it is determined that background region
Convex in screen and be in non-comfort zone;If 0 °≤ba *< 1 ° it is determined that background region convex in screen and be in comfort zone;If -1 ° <
ba *< 0 ° it is determined that background region recessed in screen and be in comfort zone;If ba *< -1 ° it is determined that background region recessed in screen and place
In non-comfort zone;
Wherein, fa *Represent the parallactic angle of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, fa *=f-k, ba *
Represent the parallactic angle of the background region in the right viewpoint parallax gray level image of pending stereo-picture, ba *=b-k, f represent human eye binocular
The convergent angle of viewing foreground target,B represents that human eye binocular watches the convergent angle of background region,K represents the adjustment angle of human eye binocular,Arctan () is just to negate
Cut function, p represents the interpupillary distance of human eye binocular, l represents the width of display, n represents the horizontal resolution of display, h represents that human eye arrives
The distance of display, f represents the mean parallax amplitude of the foreground target in the right viewpoint parallax gray level image of pending stereo-picture, f
In units of pixel,b
Represent the mean parallax amplitude of the background region in the right viewpoint parallax gray level image of pending stereo-picture, b with pixel is
Unit,dr *
(x, y) represents the pixel value of the pixel that coordinate position in the right viewpoint parallax gray level image of pending stereo-picture is (x, y);
Z2, according to the foreground target in the right viewpoint parallax gray level image of pending stereo-picture with respect to screen concavity and convexity with
And whether be in comfort zone, whether background region with respect to the concavity and convexity of screen and is in comfort zone, determines pending solid
Scene mode belonging to image, if particularly as follows: foreground target convex in screen and be in non-comfort zone, background region recessed in screen and
It is in non-comfort zone it is determined that pending stereo-picture belongs to the 1st kind of scene mode;If foreground target convex in screen and be in non-
Comfort zone, background region convex in screen and be in non-comfort zone it is determined that pending stereo-picture belongs to the 2nd kind of scene mode;
If foreground target recessed in screen and be in non-comfort zone, background region recessed in screen and be in non-comfort zone it is determined that pending
Stereo-picture belongs to the 3rd kind of scene mode;If foreground target convex in screen and be in non-comfort zone, background region recessed in screen and
It is in comfort zone it is determined that pending stereo-picture belongs to the 4th kind of scene mode;If foreground target convex in screen and be in non-relax
Suitable area, background region convex in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 5th kind of scene mode;If front
Scape target convex in screen and be in comfort zone, background region recessed in screen and be in non-comfort zone it is determined that pending stereogram
As belonging to the 6th kind of scene mode;If foreground target recessed in screen and be in comfort zone, background region recessed in screen and be in non-relaxing
Suitable area is it is determined that pending stereo-picture belongs to the 7th kind of scene mode;If foreground target convex in screen and be in comfort zone, after
Scene area recessed in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 8th kind of scene mode;If foreground target is convex
In screen and be in comfort zone, background region convex in screen and be in comfort zone it is determined that pending stereo-picture belongs to the 9th kind
Scene mode;If foreground target recessed in screen and be in comfort zone, background region recessed in screen and be in comfort zone it is determined that treating
Process stereo-picture and belong to the 10th kind of scene mode.
6. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 5, its
{ d in being characterised by described step 3.r(x, y) } in foreground target width angleWherein,
Arctan () is to negate tan, and h represents human eye to the distance of display, wfrontRepresent { dr(x, y) } in foreground target
Mean breadth, wfrontAcquisition process be:
A1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) }, by { bi
(x, y) } in coordinate position be that the pixel value of pixel of (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position be (x,
Y), when pixel belongs to foreground target, make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel belong to
When background region, make bi (x, y)=0;
A2, line scans are entered to { bi (x, y) }, for the row row in { bi (x, y) }, from the 1st pixel of row row
Start to scan to the right, when scanning the pixel that the 1st pixel value is 1, using this pixel as the 1st section of prospect width line segment
Original position, and the row coordinate of this pixel is designated as x1, continue scanning to the right until scanning the pixel that pixel value is 0
Till obtain the 1st section of prospect width line segment, the row coordinate of the pixel that this pixel value is 0 is designated as x2, by the 1st section of prospect width
The length of line segment is designated as wl1, wl1=x2-x1;Continue scan to the right, with obtain the 1st section of prospect width line segment identical mode,
Obtain all prospect width line segments in row row;Wherein, the initial value of row is 1,1≤row≤h, 1≤x1<x2≤w;
A3, length in { bi (x, y) } is less than all prospect width line segments of 0.002w and length be more than 0.995w all before
Scape width line segment removes;Then by length, order from small to large is ranked up to remaining all prospect width line segments, in taking
Between 80% prospect width line segment constitute prospect width line segment aggregate, be designated as { wl'n, wherein, wl'nRepresent { wl'nIn n-th
The length of Duan Qianjing width line segment, 1≤n≤n', n' represent { wl'nIn total hop count of prospect width line segment of comprising;
A4, calculating { wl'nIn the length of all prospect width line segments mean value, be designated as wfront',
Then by wfront' as { dr(x, y) } in foreground target mean breadth wfrontEven, wfront=wfront'.
7. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 6, its
{ d in being characterised by described step 4.r(x, y) } in foreground target average prospect line hop count{dr(x, y) } in foreground target average prospect alignment hop countIts
In, 1≤row≤h, 1≤col≤w, rlqrowRepresent { dr(x, y) } in row row in prospect line section quantity,clqcolRepresent { dr(x, y) } in col row in prospect alignment section quantity,
Wherein, rlqrowAnd clqcolAcquisition process be:
B1, to { dr(x, y) } carry out binary conversion treatment, obtain { dr(x, y) } binary image, be designated as { bi (x, y) }, by { bi
(x, y) } in coordinate position be that the pixel value of pixel of (x, y) is designated as bi (x, y), as { dr(x, y) } in coordinate position be (x,
Y), when pixel belongs to foreground target, make bi (x, y)=1, as { dr(x, y) } in coordinate position be (x, y) pixel belong to
When background region, make bi (x, y)=0;
B2,2 dilation operations that { bi (x, y) } is carried out successively with mathematical morphology, 4 erosion operations, 2 dilation operations, obtain
Arrive { bi'(x, y) }, wherein, bi'(x, y) represent that in { bi'(x, y) }, coordinate position is the pixel value of the pixel of (x, y);
The quantity of effective prospect line section in often going in b3, statistics { bi'(x, y) }, by the row row in { bi'(x, y) }
In the quantity of effective prospect line section be designated as rlqrow', rlqrow' acquisition process be: b3-1, make rlqrow' initial value be
0;B3-2, the 1st pixel of row row from { bi'(x, y) } start to scan to the right, scanning the 1st pixel value are
During 1 pixel, using this pixel as the original position of the 1st section of prospect line section, and the row coordinate of this pixel is designated as
x1', continue scanning to the right and obtain the 1st section of prospect line section till scanning the pixel that pixel value is 0, by this pixel value
The row coordinate of the pixel for 0 is designated as x2', if length x of the 1st section of prospect line section2'-x1' more than 0.005w it is determined that the 1st
Duan Qianjing line section is effective prospect line section, and makes rlqrow'=rlqrow'+1, if the length of the 1st section of prospect line section
x2'-x1' it is less than or equal to 0.005w it is determined that the 1st section of prospect line section is invalid prospect line section;Continue to scan to the right,
With with obtain the 1st section of prospect line section determine whether the 1st section of prospect line section is effective prospect line section identical mode,
Obtain all effective prospect line section in row row, and count the quantity obtaining effective prospect line section in row row
rlqrow';Wherein, the initial value of row is 1,1≤row≤h, 1≤x1'<x2'≤w, rlqrow'=rlqrowIn '+1 "=" be
Assignment;
Equally, in often going in statistics { bi'(x, y) } effective prospect alignment section quantity, by the col in { bi'(x, y) }
In row, the quantity of effective prospect alignment section is designated as clqcol', clqcol' acquisition process be: b3-1), make clqcol' initial
It is worth for 0;B3-2), the 1st pixel of the col row from { bi'(x, y) } starts to scan downwards, is scanning the 1st picture
During the pixel for 1 for the element value, using this pixel as the 1st section of prospect alignment section original position, and the row seat by this pixel
It is labeled as y1', continue scanning downwards and obtain the 1st section of prospect alignment section till scanning the pixel that pixel value is 0, should
Pixel value is that the row coordinate of 0 pixel is designated as y2', if length y of the 1st section of prospect alignment section2'-y1' it is more than 0.005h, then really
Fixed 1st section of prospect alignment section is effective prospect alignment section, and makes clqcol'=clqcol'+1, if the 1st section of prostatitis alignment section
Length y2'-y1' it is less than or equal to 0.005h it is determined that the 1st section of prospect alignment section is invalid prospect alignment section;Continue to sweep downwards
Retouch, with obtain the 1st section of prospect alignment section determine whether the 1st section of prospect alignment section is effective prospect alignment section identical side
Formula, obtains all effective prospect alignment section in col row, and counts and obtain in col row effective prospect alignment section
Quantity clqcol';Wherein, the initial value of col is 1,1≤col≤w, 1≤y1'<y2'≤h, clqcol'=clqcolIn '+1
"=" is assignment;
B4, make rlqrow=rlqrow', make clqcol=clqcol'.
8. a kind of stereo image vision comfort level evaluation method based on scene mode classification according to claim 7, its
P=1.6, t is taken in being characterised by described step 4.f=2.0, tr=2, tc=1.5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510571897.8A CN105163111B (en) | 2015-09-10 | 2015-09-10 | Method for evaluating visual comfort of three-dimensional image based on classification of scene modes |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510571897.8A CN105163111B (en) | 2015-09-10 | 2015-09-10 | Method for evaluating visual comfort of three-dimensional image based on classification of scene modes |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105163111A CN105163111A (en) | 2015-12-16 |
CN105163111B true CN105163111B (en) | 2017-01-25 |
Family
ID=54803852
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510571897.8A Active CN105163111B (en) | 2015-09-10 | 2015-09-10 | Method for evaluating visual comfort of three-dimensional image based on classification of scene modes |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105163111B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108182444A (en) * | 2017-12-08 | 2018-06-19 | 深圳英飞拓科技股份有限公司 | The method and device of video quality diagnosis based on scene classification |
CN109167988B (en) * | 2018-08-29 | 2020-06-26 | 长春理工大学 | Stereo image visual comfort evaluation method based on D + W model and contrast |
CN111860691B (en) * | 2020-07-31 | 2022-06-14 | 福州大学 | Stereo video visual comfort degree classification method based on attention and recurrent neural network |
Family Cites Families (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102724544A (en) * | 2012-06-18 | 2012-10-10 | 清华大学 | System and method for subjective assessment of stereoscopic videos |
CN103096125B (en) * | 2013-02-22 | 2015-03-04 | 吉林大学 | Stereoscopic video visual comfort evaluation method based on region segmentation |
KR101550065B1 (en) * | 2013-06-21 | 2015-09-07 | 한국과학기술원 | Stereoscopic imaging method and system for visually comfortable 3D images |
CN104853185A (en) * | 2015-06-06 | 2015-08-19 | 吉林大学 | Stereo video comfort evaluation method combining multiple parallaxes with motion |
-
2015
- 2015-09-10 CN CN201510571897.8A patent/CN105163111B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN105163111A (en) | 2015-12-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103096125B (en) | Stereoscopic video visual comfort evaluation method based on region segmentation | |
CN103581661B (en) | Method for evaluating visual comfort degree of three-dimensional image | |
CN105163111B (en) | Method for evaluating visual comfort of three-dimensional image based on classification of scene modes | |
CN106973288B (en) | A kind of three-dimensional video-frequency Comfort Evaluation method and device | |
CN104036501A (en) | Three-dimensional image quality objective evaluation method based on sparse representation | |
CN103986925B (en) | based on the stereoscopic video visual comfort evaluation method of luminance compensation | |
CN103136748B (en) | The objective evaluation method for quality of stereo images of a kind of feature based figure | |
CN106097327A (en) | In conjunction with manifold feature and the objective evaluation method for quality of stereo images of binocular characteristic | |
CN104036493B (en) | No-reference image quality evaluation method based on multifractal spectrum | |
CN105407349A (en) | No-reference objective three-dimensional image quality evaluation method based on binocular visual perception | |
CN103780895B (en) | A kind of three-dimensional video quality evaluation method | |
CN104811693A (en) | Objective evaluation method for stereo image visual comfort degree | |
CN105376563A (en) | No-reference three-dimensional image quality evaluation method based on binocular fusion feature similarity | |
CN102708567A (en) | Visual perception-based three-dimensional image quality objective evaluation method | |
CN108520510A (en) | It is a kind of based on entirety and partial analysis without referring to stereo image quality evaluation method | |
CN104902268A (en) | Non-reference three-dimensional image objective quality evaluation method based on local ternary pattern | |
CN104361583A (en) | Objective quality evaluation method of asymmetrically distorted stereo images | |
CN102708568B (en) | Stereoscopic image objective quality evaluation method on basis of structural distortion | |
CN104581141A (en) | Three-dimensional picture visual comfort evaluation method | |
CN103745457B (en) | A kind of three-dimensional image objective quality evaluation method | |
CN105898279A (en) | Stereoscopic image quality objective evaluation method | |
CN103108209B (en) | Stereo image objective quality evaluation method based on integration of visual threshold value and passage | |
CN105488792B (en) | Based on dictionary learning and machine learning without referring to stereo image quality evaluation method | |
CN104394405B (en) | A kind of method for evaluating objective quality based on full reference picture | |
CN103955921B (en) | Image noise estimation method based on human eye visual features and partitioning analysis method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20190924 Address after: 242000 Meixi Road and Wolong Lane Intersection of Ningbo Feichuan Office, Xuancheng City, Anhui Province Patentee after: Xuancheng Youdu Technology Service Co., Ltd. Address before: 315211 Zhejiang Province, Ningbo Jiangbei District Fenghua Road No. 818 Patentee before: Ningbo University |
|
TR01 | Transfer of patent right |