Computer implemented method and system for classifying an input image for new view synthesis in a 3D visual effect, and non-transitory computer readable storage medium

1. A computer implemented method for classifying an input image for a new view synthesis in a 3D visual effect, the method comprising:

creating an image representation using the input image of a size of H×W, the image representation including a set of target views based on an input 3D visual effect; and

computing an image score s t for the set of target views based on a number of non-occluded pixels in the set of target views, the image score indicating suitability of the input image for the new view synthesis in the 3D visual effect,

wherein the image representation is a multiplane image (MPI) representation defined by a set of D planar image layers, each layer encoded as an RGB-alpha image at a distance d i with respect to a viewpoint, defined by:

{( c i ,α i )} i=1 D =f θ ( I ),

where c i and α i correspond to color and alpha values of the i-th image layer, f θ represents a generic method that produces an MPI from the input image I,

wherein the MPI representation is rendered to a source viewpoint and to new viewpoints based on a warping operation and a compositing operation,

wherein the set of target views is defined by V={v 1 , v 2 , . . . , v t }, where t is a number of considered target views,

wherein the warping operation is defined by a warping function based on a depth of each image plane applied individually for color and alpha channels, as defined by:

c′ i =W v s ,v t ( d i ,c i ),

α′ i =W v s ,v t ( d i ,α i ),

where c′ i and α′ i correspond to the color and alpha values of the i-th image layer after warping from a source viewpoint v s to a target viewpoint v t , and W is a generic warping function, based on planar homography for the MPI representation,

wherein the image score is computed based on an over composite operation of the warped alpha layers, defined by:

M t =Σ i=1 D (α′ i Π j=i+1 D (1−α′ j ),

where M t is a composite alpha that represents non-occluded pixel values,

wherein the computing of the image score s t comprises calculating a metric for non-disocclusion by the following equation:

s

t

=

∑

k

=

1

H

×

W

(

M

t

(

k

)

≥

ρ

)

H

×

W

,

wherein ρ is an alpha threshold to decide whether or not an alpha value is considered as occluded or non-occluded pixel, wherein ρ is a normalized value between [0, 1].

2. The computer implemented method according to claim 1 , wherein the image representation is created from the input image and from an input depth map.

3. The computer implemented method according to claim 1 , wherein after the warping operation to the target viewpoint, a resulting MPI representation is rendered by an over composite operation, defined by:

I t =Σ i=1 D ( c′ i α′ i Π j=i+1 D (1−α′ j )),

where I t is a new rendered target view.

4. The computer implemented method according to claim 1 , wherein

the computing of the image score s t comprises counting the number of non-occluded pixels in the target views of the input image.

5. The computer implemented method according to claim 1 , wherein the image score is calculated by:

L

⁡

(

α

,

V

)

=

(

1

❘

"\[LeftBracketingBar]"

V

❘

"\[RightBracketingBar]"

⁢

∑

t

∈

V

s

t

)

⁢

min

⁡

(

s

t

⁢

❘

"\[LeftBracketingBar]"

t

∈

V

)

,

wherein V is the set of target views; s t is a metric for non-disocclusion, and

wherein L is the resulting score in a scalar value in the interval [0, 1].

6. The computer implemented method according to claim 1 , wherein the method further comprises:

establishing a threshold value for the image score, wherein the input image is suitable for a given 3D visual effect if its image score is higher than the threshold value.

7. The computer implemented method according to claim 1 , wherein the input image comprises a plurality of input images,

wherein the computing of the image score comprises computing image scores for the plurality of input images,

the method further comprising:

ranking the plurality of input images based on their corresponding image scores.

8. The computer implemented method according to claim 7 , the method further comprising:

selecting at least one best image based on the ranking of the plurality of input images, and

generating a 3D visual effect animation for the selected images.

9. The computer implemented method according to claim 7 , the method further comprising

establishing a threshold value for the image score, wherein input images are suitable for a given 3D visual effect if its imaging score is higher than the threshold value, and

generating a 3D visual effect animation for all the input images having an image score higher than the threshold value.

10. A computer implemented method for classifying an input image for a new view synthesis in a 3D visual effect, the method comprising:

creating an image representation using the input image of a size of H×W, the image representation including a set of target views based on an input 3D visual effect; and

computing an image score s t for the set of target views based on a number of non-occluded pixels in the set of target views,

wherein the image score indicates suitability of the input image for the new view synthesis in the 3D visual effect,

wherein a plurality of candidate 3D visual effects are associated with the input image, wherein for each candidate 3D visual effect the input image is used to create an image representation comprising a set of target views based on the input 3D visual effect,

wherein

the computing of the image score comprises computing image scores for the input image using each of the candidate 3D visual effects, and

ranking the plurality of candidate 3D visual effects based on corresponding image scores.

11. The computer implemented method according to claim 10 , wherein the method further comprises:

selecting a candidate 3D visual effect with a highest image score, and

generating a 3D visual effect animation selected for the input image.

12. The computer implemented method according to claim 10 , the method further comprising:

displaying for a user at least one best candidate 3D visual effect based on the ranking of the plurality of candidate 3D visual effects,

selecting, from an input of the user, a candidate 3D visual effect among the displayed at least one best candidate 3D visual effect, and

generating a 3D visual effect animation selected by the user for the input image.

13. The computer implemented method according to claim 10 , the method further comprising:

establishing a threshold value for the image score, wherein the plurality of candidate 3D visual effects are suitable for the input image if its image score is higher than the threshold value,

displaying for a user the 3D visual effect of the input image for all candidate 3D visual effects having an image score higher than the threshold value,

selecting, from the input of the user, a candidate 3D visual effect among the displayed candidate 3D visual effect, and

generating a 3D visual effect animation selected by the user for the input image.

14. The computer implemented method according to claim 10 , the method further comprising:

establishing a threshold value for the image score, wherein the candidate 3D visual effects are suitable for the input image if its image score is higher than the threshold value, and

generating 3D visual effect animations for all the suitable candidate 3D visual effects for the input image.

15. A system for classifying an input image for new view synthesis in a 3D visual effect, comprising:

a processor; and

a memory including computer readable instructions that, when executed by the processor, causes the processor to perform the method as defined in claim 10 .

16. A non-transitory computer readable storage medium which stores computer readable instructions that, when executed by a processor, causes the processor to perform the method as defined in claim 10 .