Summary of the invention
Technical problem to be solved by this invention is to provide a kind of objective evaluation method for quality of stereo images based on picture breakdown that effectively can improve the correlation of objective evaluation result and subjective perception.
The present invention solves the problems of the technologies described above adopted technical scheme: a kind of objective evaluation method for quality of stereo images based on picture breakdown, it is characterized in that its processing procedure is:
First, 3 grades of wavelet transformations are implemented to the left visual point image of the left visual point image of original undistorted stereo-picture and the stereo-picture of distortion to be evaluated, again according to the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation obtaining, obtain Recovery image and the interfering picture of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, 3 grades of wavelet transformations are implemented to the right visual point image of the right visual point image of original undistorted stereo-picture and the stereo-picture of distortion to be evaluated, again according to the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation obtaining, obtain Recovery image and the interfering picture of the right visual point image of the stereo-picture of distortion to be evaluated;
Secondly, by the local phase characteristic sum local amplitude feature of each pixel in the Recovery image of the left visual point image of the stereo-picture of the left visual point image and distortion to be evaluated that calculate original undistorted stereo-picture, obtain the picture quality objective evaluation predicted value of the Recovery image of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, by the local phase characteristic sum local amplitude feature of each pixel in the Recovery image of the right visual point image of the stereo-picture of the right visual point image and distortion to be evaluated that calculate original undistorted stereo-picture, obtain the picture quality objective evaluation predicted value of the Recovery image of the right visual point image of the stereo-picture of distortion to be evaluated;
Then, by the singular value vector that each size in the interfering picture of the left visual point image of the stereo-picture of the left visual point image and distortion to be evaluated that calculate original undistorted stereo-picture is the sub-block of 8 × 8, obtain the picture quality objective evaluation predicted value of the interfering picture of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, by the singular value vector that each size in the interfering picture of the right visual point image of the stereo-picture of the right visual point image and distortion to be evaluated that calculate original undistorted stereo-picture is the sub-block of 8 × 8, obtain the picture quality objective evaluation predicted value of the interfering picture of the right visual point image of the stereo-picture of distortion to be evaluated;
Afterwards, the Recovery image of the left visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of interfering picture are merged, obtains the picture quality objective evaluation predicted value of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, the Recovery image of the right visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of interfering picture are merged, obtains the picture quality objective evaluation predicted value of the right visual point image of the stereo-picture of distortion to be evaluated;
Moreover, the left visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of right visual point image are merged, obtains the picture quality objective evaluation predicted value of the stereo-picture of distortion to be evaluated;
Finally, adopt the undistorted stereo-picture that several are original, set up its distortion stereo-picture set under the different distortion level of different type of distortion, this distortion stereo-picture set comprises the stereo-picture of several distortions, then calculates the picture quality objective evaluation predicted value of the stereo-picture of every width distortion in this distortion stereo-picture set respectively according to the process of the picture quality objective evaluation predicted value of the stereo-picture of above-mentioned acquisition distortion to be evaluated.
Objective evaluation method for quality of stereo images based on picture breakdown of the present invention, is characterized in that it specifically comprises the following steps:
1. S is made
orgfor original undistorted stereo-picture, make S
disfor the stereo-picture of distortion to be evaluated, by S
orgleft visual point image be designated as { L
org(x, y) }, by S
orgright visual point image be designated as { R
org(x, y) }, by S
disleft visual point image be designated as { L
dis(x, y) }, by S
disright visual point image be designated as { R
dis(x, y) }, wherein, (x, y) coordinate position of the pixel in left visual point image and right visual point image is represented, 1≤x≤W, 1≤y≤H, W represents the width of left visual point image and right visual point image, and H represents the height of left visual point image and right visual point image, L
org(x, y) represents { L
org(x, y) } in coordinate position be the pixel value of the pixel of (x, y), R
org(x, y) represents { R
org(x, y) } in coordinate position be the pixel value of the pixel of (x, y), L
dis(x, y) represents { L
dis(x, y) } in coordinate position be the pixel value of the pixel of (x, y), R
dis(x, y) represents { R
dis(x, y) } in coordinate position be the pixel value of the pixel of (x, y);
2. respectively to { L
org(x, y) } and { L
dis(x, y) } implement 3 grades of wavelet transformations, then according to { L
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { L
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, obtain { L
dis(x, y) } Recovery image and interfering picture, correspondence is designated as
with
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y),
represent
middle coordinate position is the pixel value of the pixel of (x, y);
Respectively to { R
org(x, y) } and { R
dis(x, y) } implement 3 grades of wavelet transformations, then according to { R
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { R
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, obtain { R
dis(x, y) } Recovery image and interfering picture, correspondence is designated as
with
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y),
represent
middle coordinate position is the pixel value of the pixel of (x, y);
3. { L is calculated respectively
org(x, y) }, { R
org(x, y) },
with
in the local phase characteristic sum local amplitude feature of each pixel, by { L
org(x, y) } in coordinate position be that the local phase feature of the pixel of (x, y) is designated as
by { L
org(x, y) } in coordinate position be that the local amplitude feature of the pixel of (x, y) is designated as
by { R
org(x, y) } in coordinate position be that the local phase feature of the pixel of (x, y) is designated as
by { R
org(x, y) } in coordinate position be that the local amplitude feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local phase feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local amplitude feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local phase feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local amplitude feature of the pixel of (x, y) is designated as
4. according to { L
org(x, y) } and
in the local phase characteristic sum local amplitude feature of each pixel, calculate
picture quality objective evaluation predicted value, be designated as
Wherein, T
1and T
2for controling parameters;
According to { R
org(x, y) } and
in the local phase characteristic sum local amplitude feature of each pixel, calculate
picture quality objective evaluation predicted value, be designated as
Wherein, T
1and T
2for controling parameters;
5. respectively by { L
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, then respectively to { L
org(x, y) } in each sub-block and
in each sub-block implement singular value decomposition, obtain { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, then to calculate
picture quality objective evaluation predicted value, be designated as
wherein, N
blockrepresent { L
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, " <> " for getting interior Product function,
represent { L
org(x, y) } in the singular value vector of a kth sub-block,
represent
in the singular value vector of a kth sub-block;
Respectively by { R
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, then respectively to { R
org(x, y) } in each sub-block and
in each sub-block implement singular value decomposition, obtain { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, then to calculate
picture quality objective evaluation predicted value, be designated as
wherein, N '
blockshow { R
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, " <> " for getting interior Product function,
represent { R
org(x, y) } in the singular value vector of a kth sub-block,
represent
in the singular value vector of a kth sub-block;
6. right
picture quality objective evaluation predicted value
with
picture quality objective evaluation predicted value
merge, obtain { L
dis(x, y) } picture quality objective evaluation predicted value, be designated as Q
l,
and it is right
picture quality objective evaluation predicted value
with
picture quality objective evaluation predicted value
merge, obtain { R
dis(x, y) } picture quality objective evaluation predicted value, be designated as Q
r,
wherein, w
1represent
with
weights proportion, w
2represent
with
weights proportion, w
1+ w
2=1;
7. to { L
dis(x, y) } picture quality objective evaluation predicted value Q
l{ R
dis(x, y) } picture quality objective evaluation predicted value Q
rmerge, obtain S
dispicture quality objective evaluation predicted value, be designated as Q,
Wherein, p, q, m, s and z are model coefficient.
Described step detailed process is 2.:
2.-1, to { L
org(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { L
org(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
wherein, 3 directional subbands are respectively horizontal direction subband, vertical direction subband and diagonal angle directional subband, 1≤m≤3,1≤n≤3,
represent
middle coordinate position is the wavelet coefficient at (x, y) place;
2.-2, to { L
dis(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
wherein, 3 directional subbands are respectively horizontal direction subband, vertical direction subband and diagonal angle directional subband, 1≤m≤3,1≤n≤3,
represent
middle coordinate position is the wavelet coefficient at (x, y) place;
2.-3, according to { L
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { L
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, estimate to obtain { L
dis(x, y) } the matrix of wavelet coefficients compensating parameter matrix separately of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
compensating parameter matrix be designated as
wherein,
represent
middle coordinate position is the compensating parameter at (x, y) place,
for inputting
be truncated to the truncation funcation that [0,1] is interval;
2.-4, according to { L
org(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients compensating parameter matrix separately of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, calculate { L
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after recovering is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
then to { L
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding implements anti-wavelet transformation, obtains { L
dis(x, y) } Recovery image, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y);
2.-5, according to { L
org(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } matrix of wavelet coefficients that the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding obtains after hanging oneself and recovering, calculate { L
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding to hang oneself the matrix of wavelet coefficients obtained after interference, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after interference is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
Then to { L
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations the is corresponding matrix of wavelet coefficients obtained after interference of hanging oneself implements anti-wavelet transformation, obtains { L
dis(x, y) } interfering picture, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y);
2.-6,2.-1 2.-5 { L are obtained to step according to step
dis(x, y) } Recovery image
{ L
dis(x, y) } interfering picture
operation, in an identical manner obtain { R
dis(x, y) } Recovery image
{ R
dis(x, y) } interfering picture
Described step detailed process is 3.:
3.-1, adopt log-Garbor filter to { L
org(x, y) } in each pixel carry out filtering process, obtain { L
org(x, y) } in each pixel in the even symmetry frequency response of different scale and different directions and odd symmetry frequency response, by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as e in the even symmetry frequency response of different scale and different directions
α, θ(x, y), by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as o in the odd symmetry frequency response of different scale and different directions
α, θ(x, y), wherein, α represents the scale factor of log-Garbor filter, 1≤α≤4, and θ represents the direction factor of log-Garbor filter, 1≤θ≤4;
3.-2, { L is calculated
org(x, y) } in each pixel in the phase equalization feature of different directions, by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as PC in the phase equalization feature of different directions
θ(x, y),
Wherein,
3.-3, according to { L
org(x, y) } in direction corresponding to the maximum phase consistency feature of each pixel, calculate { L
org(x, y) } in the local phase characteristic sum local amplitude feature of each pixel, for { L
org(x, y) } in coordinate position be the pixel of (x, y), first find out its maximum phase consistency feature in the phase equalization feature of different directions, next finds out direction corresponding to this maximum phase consistency feature, is designated as θ
m, again according to θ
mcalculate { L
org(x, y) } in coordinate position be the local phase characteristic sum local amplitude feature of the pixel of (x, y), correspondence is designated as
with
Wherein,
represent { L
org(x, y) } in coordinate position be that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
meven symmetry frequency response,
represent
middle coordinate position is that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
modd symmetry frequency response, arctan () is negate tan;
3.-4,3.-1 3.-3 { L are obtained to step according to step
org(x, y) } in the operation of local phase characteristic sum local amplitude feature of each pixel, obtain { R in an identical manner
org(x, y) },
with
in the local phase characteristic sum local amplitude feature of each pixel.
Described step 5. in
picture quality objective evaluation predicted value
acquisition process be:
5.-1a, respectively by { L
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, by { L
org(x, y) } in a current pending kth sub-block be defined as current first sub-block, will
in the sub-block of current pending kth be defined as current second sub-block, wherein,
5.-2a, current first sub-block is designated as
current second sub-block is designated as
wherein, (x
2, y
2) represent
with
in the coordinate position of pixel, 1≤x
2≤ 8,1≤y
2≤ 8,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel;
5.-3a, general
be expressed as in the form of vectors
right
implement singular value decomposition,
Wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector;
Will
be expressed as in the form of vectors
right
implement singular value decomposition,
wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector;
5.-4a, make k=k+1, by { L
org(x, y) } in next pending sub-block as current first sub-block, will
the pending sub-block of the middle next one as current second sub-block, then return step 5.-2a continue to perform, until { L
org(x, y) } and
in all sub-blocks be all disposed, obtain { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, wherein, "=" in k=k+1 is assignment;
5.-5a, basis { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, calculate
picture quality objective evaluation predicted value, be designated as
wherein, N
blockrepresent { L
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, and " <> " is for getting interior Product function;
Described step 5. in
picture quality objective evaluation predicted value
acquisition process be:
5.-1b, respectively by { R
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, by { R
org(x, y) } in a current pending kth sub-block be defined as current first sub-block, will
in the sub-block of current pending kth be defined as current second sub-block, wherein,
5.-2b, current first sub-block is designated as
current second sub-block is designated as
wherein, (x
2, y
2) represent
with
in the coordinate position of pixel, 1≤x
2≤ 8,1≤y
2≤ 8,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel;
5.-3b, general
be expressed as in the form of vectors
right
implement singular value decomposition,
Wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector;
Will
be expressed as in the form of vectors
right
implement singular value decomposition,
wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector;
5.-4b, make k=k+1, by { R
org(x, y) } in next pending sub-block as current first sub-block, will
the pending sub-block of the middle next one as current second sub-block, then return step 5.-2b continue to perform, until { R
org(x, y) } and
in all sub-blocks be all disposed, obtain { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, wherein, "=" in k=k+1 is assignment;
5.-5b, basis { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, calculate
picture quality objective evaluation predicted value, be designated as
wherein, N '
blcokrepresent { R
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, and " <> " is for getting interior Product function.
Described step 4. in get T
1=0.85, T
2=160.
Described step 6. in get w
1=0.9208, w
2=0.0792.
Described step 7. in get p=7.99, q=6.59, m=1.28, s=0.985 and z=0.077.
Compared with prior art, the invention has the advantages that:
1) the inventive method considers that distortion can cause image detail to be lost or redundant information increases, therefore the stereo-picture of distortion is decomposed into Recovery image and interfering picture, and respectively Recovery image and interfering picture are evaluated, the mass change situation of stereo-picture can be reflected so preferably, make evaluation result more meet human visual system.
2) the inventive method adopts local phase characteristic sum local amplitude feature to evaluate Recovery image, singular value vector is adopted to evaluate interfering picture, token image details and redundant information on the impact of picture quality, thus can effectively can improve the correlation of objective evaluation result and subjective perception well like this.
Embodiment
Below in conjunction with accompanying drawing embodiment, the present invention is described in further detail.
A kind of objective evaluation method for quality of stereo images based on picture breakdown that the present invention proposes, it totally realizes block diagram as shown in Figure 1, and its processing procedure is:
First, 3 grades of wavelet transformations are implemented to the left visual point image of the left visual point image of original undistorted stereo-picture and the stereo-picture of distortion to be evaluated, again according to the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation obtaining, obtain Recovery image and the interfering picture of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, 3 grades of wavelet transformations are implemented to the right visual point image of the right visual point image of original undistorted stereo-picture and the stereo-picture of distortion to be evaluated, again according to the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation obtaining, obtain Recovery image and the interfering picture of the right visual point image of the stereo-picture of distortion to be evaluated;
Secondly, by the local phase characteristic sum local amplitude feature of each pixel in the Recovery image of the left visual point image of the stereo-picture of the left visual point image and distortion to be evaluated that calculate original undistorted stereo-picture, obtain the picture quality objective evaluation predicted value of the Recovery image of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, by the local phase characteristic sum local amplitude feature of each pixel in the Recovery image of the right visual point image of the stereo-picture of the right visual point image and distortion to be evaluated that calculate original undistorted stereo-picture, obtain the picture quality objective evaluation predicted value of the Recovery image of the right visual point image of the stereo-picture of distortion to be evaluated;
Then, by the singular value vector that each size in the interfering picture of the left visual point image of the stereo-picture of the left visual point image and distortion to be evaluated that calculate original undistorted stereo-picture is the sub-block of 8 × 8, obtain the picture quality objective evaluation predicted value of the interfering picture of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, by the singular value vector that each size in the interfering picture of the right visual point image of the stereo-picture of the right visual point image and distortion to be evaluated that calculate original undistorted stereo-picture is the sub-block of 8 × 8, obtain the picture quality objective evaluation predicted value of the interfering picture of the right visual point image of the stereo-picture of distortion to be evaluated;
Afterwards, the Recovery image of the left visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of interfering picture are merged, obtains the picture quality objective evaluation predicted value of the left visual point image of the stereo-picture of distortion to be evaluated; Equally, the Recovery image of the right visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of interfering picture are merged, obtains the picture quality objective evaluation predicted value of the right visual point image of the stereo-picture of distortion to be evaluated;
Moreover, the left visual point image of the stereo-picture of distortion to be evaluated and the picture quality objective evaluation predicted value of right visual point image are merged, obtains the picture quality objective evaluation predicted value of the stereo-picture of distortion to be evaluated;
Finally, adopt the undistorted stereo-picture that several are original, set up its distortion stereo-picture set under the different distortion level of different type of distortion, this distortion stereo-picture set comprises the stereo-picture of several distortions, then calculates the picture quality objective evaluation predicted value of the stereo-picture of every width distortion in this distortion stereo-picture set respectively according to the process of the picture quality objective evaluation predicted value of the stereo-picture of above-mentioned acquisition distortion to be evaluated.
Objective evaluation method for quality of stereo images of the present invention, specifically comprises the following steps:
1. S is made
orgfor original undistorted stereo-picture, make S
disfor the stereo-picture of distortion to be evaluated, by S
orgleft visual point image be designated as { L
org(x, y) }, by S
orgright visual point image be designated as { R
org(x, y) }, by S
disleft visual point image be designated as { L
dis(x, y) }, by S
disright visual point image be designated as { R
dis(x, y) }, wherein, (x, y) coordinate position of the pixel in left visual point image and right visual point image is represented, 1≤x≤W, 1≤y≤H, W represents the width of left visual point image and right visual point image, and H represents the height of left visual point image and right visual point image, L
org(x, y) represents { L
org(x, y) } in coordinate position be the pixel value of the pixel of (x, y), R
org(x, y) represents { R
org(x, y) } in coordinate position be the pixel value of the pixel of (x, y), L
dis(x, y) represents { L
dis(x, y) } in coordinate position be the pixel value of the pixel of (x, y), R
dis(x, y) represents { R
dis(x, y) } in coordinate position be the pixel value of the pixel of (x, y).
2. can distinguish due to distortion and introduce information dropout distortion (information-loss distortion) and information interpolation distortion (information-additive distortion) in the picture, two kinds of distortion impacts on perceived quality are different, such as information dropout distortion can cause binocular to suppress, and the impact of information interpolation distortion on perception is not very large, therefore the inventive method is respectively to { L
org(x, y) } and { L
dis(x, y) } implement 3 grades of wavelet transformations, then according to { L
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { L
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, obtain { L
dis(x, y) } Recovery image and interfering picture, correspondence is designated as
with
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y),
represent
middle coordinate position is the pixel value of the pixel of (x, y); Respectively to { R
org(x, y) } and { R
dis(x, y) } implement 3 grades of wavelet transformations, then according to { R
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { R
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, obtain { R
dis(x, y) } Recovery image and interfering picture, correspondence is designated as
with
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y),
represent
middle coordinate position is the pixel value of the pixel of (x, y).
In this particular embodiment, step detailed process is 2.:
2.-1, to { L
org(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { L
org(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
wherein, 3 directional subbands are respectively horizontal direction subband, vertical direction subband and diagonal angle directional subband, 1≤m≤3,1≤n≤3, during n=1, the n-th directional subband is horizontal direction subband, and during n=2, the n-th directional subband is vertical direction subband, during n=3, the n-th directional subband is diagonal angle directional subband
represent
middle coordinate position is the wavelet coefficient at (x, y) place.
2.-2, to { L
dis(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
wherein, 3 directional subbands are respectively horizontal direction subband, vertical direction subband and diagonal angle directional subband, 1≤m≤3,1≤n≤3, during n=1, the n-th directional subband is horizontal direction subband, and during n=2, the n-th directional subband is vertical direction subband, during n=3, the n-th directional subband is diagonal angle directional subband
represent
middle coordinate position is the wavelet coefficient at (x, y) place.
2.-3, according to { L
org(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding and { L
dis(x, y) } matrix of wavelet coefficients of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, estimate to obtain { L
dis(x, y) } the matrix of wavelet coefficients compensating parameter matrix separately of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
compensating parameter matrix be designated as
wherein,
represent
middle coordinate position is the compensating parameter at (x, y) place, namely represents
middle coordinate position is the wavelet coefficient at (x, y) place
compensating parameter,
for inputting
be truncated to the truncation funcation that [0,1] is interval.
2.-4, according to { L
org(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients compensating parameter matrix separately of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, calculate { L
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after recovering is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
then to { L
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding implements anti-wavelet transformation, obtains { L
dis(x, y) } Recovery image, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y).
2.-5, according to { L
org(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } the matrix of wavelet coefficients, { L of 3 directional subbands that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding
dis(x, y) } matrix of wavelet coefficients that the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding obtains after hanging oneself and recovering, calculate { L
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding to hang oneself the matrix of wavelet coefficients obtained after interference, by { L
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after interference is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
Then to { L
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations the is corresponding matrix of wavelet coefficients obtained after interference of hanging oneself implements anti-wavelet transformation, obtains { L
dis(x, y) } interfering picture, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y).
2.-6,2.-1 2.-5 { L are obtained to step according to step
dis(x, y) } Recovery image
{ L
dis(x, y) } interfering picture
operation, in an identical manner obtain { R
dis(x, y) } Recovery image
{ R
dis(x, y) } interfering picture
namely detailed process is: 1) to { R
org(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { R
org(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
represent
middle coordinate position is the wavelet coefficient at (x, y) place; 2) to { R
dis(x, y) } implement 3 grades of wavelet transformations, obtain the matrix of wavelet coefficients of 3 directional subbands corresponding to every grade of wavelet transformation, by { R
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains be designated as
represent
middle coordinate position is the wavelet coefficient at (x, y) place; 3) { R is calculated
dis(x, y) } the matrix of wavelet coefficients compensating parameter matrix separately of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { R
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
compensating parameter matrix be designated as
wherein
represent
middle coordinate position is the compensating parameter at (x, y) place,
4) { R is calculated
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding, by { R
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after recovering is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
then to { R
dis(x, y) } matrix of wavelet coefficients that obtains after hanging oneself and recovering of the matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding implements anti-wavelet transformation, obtains { R
dis(x, y) } Recovery image, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y); 5) { R is calculated
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations is corresponding to hang oneself the matrix of wavelet coefficients obtained after interference, by { R
dis(x, y) } implement m level wavelet transformation after the matrix of wavelet coefficients of the n-th directional subband that obtains
the matrix of wavelet coefficients obtained after interference is designated as
wherein,
represent
middle coordinate position is the wavelet coefficient at (x, y) place,
Then to { R
dis(x, y) } matrix of wavelet coefficients of all directions subband that every grade of wavelet transformation implementing to obtain after 3 grades of wavelet transformations the is corresponding matrix of wavelet coefficients obtained after interference of hanging oneself implements anti-wavelet transformation, obtains { R
dis(x, y) } interfering picture, be designated as
wherein,
represent
middle coordinate position is the pixel value of the pixel of (x, y).
3. { L is calculated respectively
org(x, y) }, { R
org(x, y) },
with
in the local phase characteristic sum local amplitude feature of each pixel, by { L
org(x, y) } in coordinate position be that the local phase feature of the pixel of (x, y) is designated as
by { L
org(x, y) } in coordinate position be that the local amplitude feature of the pixel of (x, y) is designated as
by { R
org(x, y) } in coordinate position be that the local phase feature of the pixel of (x, y) is designated as
by { R
org(x, y) } in coordinate position be that the local amplitude feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local phase feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local amplitude feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local phase feature of the pixel of (x, y) is designated as
will
middle coordinate position is that the local amplitude feature of the pixel of (x, y) is designated as
In this particular embodiment, step detailed process is 3.:
3.-1, adopt log-Garbor filter to { L
org(x, y) } in each pixel carry out filtering process, obtain { L
org(x, y) } in each pixel in the even symmetry frequency response of different scale and different directions and odd symmetry frequency response, by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as e in the even symmetry frequency response of different scale and different directions
α, θ(x, y), by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as o in the odd symmetry frequency response of different scale and different directions
α, θ(x, y), wherein, α represents the scale factor of log-Garbor filter, 1≤α≤4, and θ represents the direction factor of log-Garbor filter, 1≤θ≤4.
3.-2, { L is calculated
org(x, y) } in each pixel in the phase equalization feature of different directions, by { L
org(x, y) } in coordinate position be that the pixel of (x, y) is designated as PC in the phase equalization feature of different directions
θ(x, y),
Wherein,
3.-3, according to { L
org(x, y) } in direction corresponding to the maximum phase consistency feature of each pixel, calculate { L
org(x, y) } in the local phase characteristic sum local amplitude feature of each pixel, for { L
org(x, y) } in coordinate position be the pixel of (x, y), first find out its maximum phase consistency feature in the phase equalization feature of different directions, next finds out direction corresponding to this maximum phase consistency feature, is designated as θ
m, again according to θ
mcalculate { L
org(x, y) } in coordinate position be the local phase characteristic sum local amplitude feature of the pixel of (x, y), correspondence is designated as
with
Wherein,
represent { L
org(x, y) } in coordinate position be that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
meven symmetry frequency response,
represent { L
org(x, y) } in coordinate position be that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
modd symmetry frequency response, arctan () is negate tan.
3.-4,3.-1 3.-3 { L are obtained to step according to step
org(x, y) } in the operation of local phase characteristic sum local amplitude feature of each pixel, obtain { R in an identical manner
org(x, y) },
with
in the local phase characteristic sum local amplitude feature of each pixel.As: obtain
in the detailed process of local phase characteristic sum local amplitude feature of each pixel be: 1) adopt log-Garbor filter pair
in each pixel carry out filtering process, obtain
in each pixel in the even symmetry frequency response of different scale and different directions and odd symmetry frequency response, will
middle coordinate position is that the pixel of (x, y) is designated as e' in the even symmetry frequency response of different scale and different directions
α, θ(x, y), will
middle coordinate position is that the pixel of (x, y) is designated as o' in the odd symmetry frequency response of different scale and different directions
α, θ(x, y); 2) calculate
in each pixel in the phase equalization feature of different directions, will
middle coordinate position is that the pixel of (x, y) is designated as PC' in the phase equalization feature of different directions
θ(x, y),
wherein,
3) basis
in direction corresponding to the maximum phase consistency feature of each pixel, calculate
in the local phase characteristic sum local amplitude feature of each pixel, for
middle coordinate position is the pixel of (x, y), and first find out its maximum phase consistency feature in the phase equalization feature of different directions, next finds out direction corresponding to this maximum phase consistency feature, is designated as θ
m, again according to θ
mcalculate
middle coordinate position is the local phase characteristic sum local amplitude feature of the pixel of (x, y), and correspondence is designated as
with
Wherein,
represent
middle coordinate position is that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
meven symmetry frequency response,
represent
middle coordinate position is that the pixel of (x, y) is at different scale and direction θ corresponding to maximum phase consistency feature
modd symmetry frequency response.
4. compared with original image, Recovery image can introduce information dropout distortion (information-loss distortion), and local phase characteristic sum local amplitude feature can evaluate image detail information change well, and therefore the inventive method is according to { L
org(x, y) } and
in the local phase characteristic sum local amplitude feature of each pixel, calculate
picture quality objective evaluation predicted value, be designated as
Wherein, T
1and T
2for controling parameters; And according to
with
in the local phase characteristic sum local amplitude feature of each pixel, calculate
picture quality objective evaluation predicted value, be designated as
Wherein, T
1and T
2for controling parameters.In the present embodiment, T is got
1=0.85, T
2=160.
5. interfering picture can be introduced information and adds distortion (information-additive distortion), and singular value can Description Image energy well, and the change of singular value can be utilized to reflect redundant information, and therefore the inventive method is respectively by { L
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, then respectively to { L
org(x, y) } in each sub-block and
in each sub-block implement singular value decomposition, obtain { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, then to calculate
picture quality objective evaluation predicted value, be designated as
Wherein, N
blockrepresent { L
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, " <> " for getting interior Product function,
represent { L
org(x, y) } in the singular value vector of a kth sub-block,
represent
in the singular value vector of a kth sub-block; Equally, respectively will
with
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, then respectively to { R
org(x, y) } in each sub-block and
in each sub-block implement singular value decomposition, obtain { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, then to calculate
picture quality objective evaluation predicted value, be designated as
wherein, N '
blockrepresent { R
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, " <> " for getting interior Product function,
represent
in the singular value vector of a kth sub-block,
represent
in the singular value vector of a kth sub-block.
In this particular embodiment, step 5. in
picture quality objective evaluation predicted value
acquisition process be:
5.-1a, respectively by { L
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, by { L
org(x, y) } in a current pending kth sub-block be defined as current first sub-block, will
in the sub-block of current pending kth be defined as current second sub-block, wherein,
5.-2a, current first sub-block is designated as
current second sub-block is designated as
wherein, (x
2, y
2) represent
with
in the coordinate position of pixel, 1≤x
2≤ 8,1≤y
2≤ 8,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel.
5.-3a, general
be expressed as in the form of vectors
right
implement singular value decomposition,
Wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector; Will
be expressed as in the form of vectors
right
implement singular value decomposition,
wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector.
5.-4a, make k=k+1, by { L
org(x, y) } in next pending sub-block as current first sub-block, will
the pending sub-block of the middle next one as current second sub-block, then return step 5.-2a continue to perform, until { L
org(x, y) } and
in all sub-blocks be all disposed, obtain { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, wherein, "=" in k=k+1 is assignment.
5.-5a, basis { L
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, calculate
picture quality objective evaluation predicted value, be designated as
wherein, N
blockrepresent { L
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, and " <> " is for getting interior Product function.
In this particular embodiment, step 5. in
picture quality objective evaluation predicted value
acquisition process be:
5.-1b, respectively by { R
org(x, y) } and
be divided into
the size of individual non-overlapping copies is the sub-block of 8 × 8, by { R
org(x, y) } in a current pending kth sub-block be defined as current first sub-block, will
in the sub-block of current pending kth be defined as current second sub-block, wherein,
5.-2b, current first sub-block is designated as
current second sub-block is designated as
wherein, (x
2, y
2) represent
with
in the coordinate position of pixel, 1≤x
2≤ 8,1≤y
2≤ 8,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel,
represent
middle coordinate position is (x
2, y
2) the pixel value of pixel.
5.-3b, general
be expressed as in the form of vectors
right
implement singular value decomposition,
Wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector; Will
be expressed as in the form of vectors
right
implement singular value decomposition,
wherein,
for
left singular vector,
for
right singular vector,
for
singular value vector,
diagonal on element be singular value, and its value arranges according to order from big to small,
for
transposed vector.
5.-4b, make k=k+1, by { R
org(x, y) } in next pending sub-block as current first sub-block, will
the pending sub-block of the middle next one as current second sub-block, then return step 5.-2b continue to perform, until { R
org(x, y) } and
in all sub-blocks be all disposed, obtain { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, wherein, "=" in k=k+1 is assignment.
5.-5b, basis { R
org(x, y) } in each sub-block singular value vector and
in the singular value vector of each sub-block, calculate
picture quality objective evaluation predicted value, be designated as
wherein, N '
blockrepresent { R
org(x, y) } in the number of sub-block that comprises, also represent
in the number of sub-block that comprises,
symbol " || " is the symbol that takes absolute value, and " <> " is for getting interior Product function.
6. right
picture quality objective evaluation predicted value
with
picture quality objective evaluation predicted value
merge, obtain { L
dis(x, y) } picture quality objective evaluation predicted value, be designated as Q
l,
and it is right
picture quality objective evaluation predicted value
with
picture quality objective evaluation predicted value
merge, obtain { R
dis(x, y) } picture quality objective evaluation predicted value, be designated as
wherein, w
1represent
with
weights proportion, w
2represent
with
weights proportion, w
1+ w
2=1, in the present embodiment, get w
1=0.9208, w
2=0.0792.
7. to { L
dis(x, y) } picture quality objective evaluation predicted value Q
l{ R
dis(x, y) } picture quality objective evaluation predicted value Q
rmerge, obtain S
dispicture quality objective evaluation predicted value, be designated as Q,
Wherein, p, q, m, s and z are model coefficient, in the present embodiment, get p=7.99, q=6.59, m=1.28, s=0.985 and z=0.077.
8. the undistorted stereo-picture that n original is adopted, set up its distortion stereo-picture set under the different distortion level of different type of distortion, this distortion stereo-picture set comprises the stereo-picture of several distortions, utilizes subjective quality assessment method to obtain the mean subjective scoring difference of the stereo-picture of every width distortion in this distortion stereo-picture set respectively, is designated as DMOS, DMOS=100-MOS, wherein, MOS represents subjective scoring average, DMOS ∈ [0,100], n>=1; Then 1. 7. S is calculated to step according to step
disthe operation of picture quality objective evaluation predicted value Q, calculate the picture quality objective evaluation predicted value of the stereo-picture of every width distortion in this distortion stereo-picture set in an identical manner respectively.
In the present embodiment, utilize the stereo-picture as Fig. 2 a and Fig. 2 b is formed, the stereo-picture that Fig. 3 a and Fig. 3 b is formed, the stereo-picture that Fig. 4 a and Fig. 4 b is formed, the stereo-picture that Fig. 5 a and Fig. 5 b is formed, the stereo-picture that Fig. 6 a and Fig. 6 b is formed, the stereo-picture that Fig. 7 a and Fig. 7 b is formed, the stereo-picture that Fig. 8 a and Fig. 8 b is formed, the stereo-picture that Fig. 9 a and Fig. 9 b is formed, stereo-picture totally 9 width (n=9) the undistorted stereo-picture that Figure 10 a and Figure 10 b is formed, set up correspondence 5 specified distortion level under Gaussian Blur, lower 5 specified distortion level of white Gaussian noise, JPEG compresses lower 5 specified distortion level, JPEG2000 compresses lower 5 specified distortion level, H.264 the 234 width distortion stereo-pictures altogether of lower 6 specified distortion level are compressed as test stereo-picture.This 234 width distortion stereo-picture forms the set of a distortion stereo-picture, existing subjective quality assessment method is utilized to obtain the mean subjective scoring difference of the stereo-picture of every width distortion in this distortion stereo-picture set respectively, be designated as DMOS, DMOS=100-MOS, wherein, MOS represents subjective scoring average, DMOS ∈ [0,100]; Then 1. 7. S is calculated to step according to step
disthe operation of picture quality objective evaluation predicted value Q, calculate the picture quality objective evaluation predicted value of the stereo-picture of every width distortion in this distortion stereo-picture set in an identical manner respectively.
9 undistorted stereo-pictures shown in Fig. 2 a to Figure 10 b are adopted to analyze at the stereo-picture of JPEG compression in various degree, JPEG2000 compression, Gaussian Blur, white noise and 234 width distortions H.264 in coding distortion situation the correlation that the picture quality objective evaluation predicted value of the stereo-picture of this 234 width distortion and mean subjective mark between difference.In the present embodiment, utilize 4 of evaluate image quality evaluating method conventional objective parameters as evaluation index, namely Pearson correlation coefficient (the Pearson linear correlation coefficient under nonlinear regression condition, PLCC), Spearman coefficient correlation (Spearman rank order correlation coefficient, SROCC), Kendall coefficient correlation (Kendall rank-order correlation coefficient, KROCC), mean square error (root mean squarederror, RMSE), PLCC and RMSE reflects the accuracy of the picture quality objective evaluation predicted value of the stereo-picture of distortion, SROCC and KROCC reflects its monotonicity.The picture quality objective evaluation predicted value of the stereo-picture of the 234 width distortions calculated is done four parameter Logistic function nonlinear fittings, PLCC and SROCC value is higher, the less explanation of OR and RMSE value assessment method for encoding quality of the present invention and mean subjective difference correlation of marking is better.PLCC, SROCC, KROCC and RMSE coefficient of reflection three-dimensional image objective evaluation method performance is as shown in table 1, from the data listed by table 1, final picture quality objective evaluation predicted value and the mean subjective correlation of marking between difference of the stereo-picture of the distortion obtained by the inventive method are very high, the result fully indicating objective evaluation result and human eye subjective perception is more consistent, is enough to the validity that the inventive method is described.
Figure 11 gives the scatter diagram that the picture quality objective evaluation predicted value of the stereo-picture of 234 width distortions and mean subjective mark difference, and loose point is more concentrated, illustrates that the consistency of objective evaluation result and subjective perception is better.As can be seen from Figure 11, adopt the scatter diagram that obtains of the inventive method more concentrated, and the goodness of fit between subjective assessment data is higher.
Correlation between the picture quality objective evaluation predicted value of the stereo-picture of the 234 width distortions that table 1 utilizes the inventive method to obtain and subjective scoring