Method and device for determining scoring models of a three-dimensional animation scene frame

Patent metadata

Field	Value
Publication number	US-9965884-B2
Application number	US-201614993612-A
Country	US
Kind code	B2
Filing date	Jan 12, 2016
Priority date	Oct 15, 2015
Publication date	May 8, 2018
Grant date	May 8, 2018

Abstract

Methods and devices for determining scoring models of a three-dimensional animation scene frame are provided. A method can include obtaining a dataset of three-dimensional animation scene frames; obtaining a predetermined stereoscopic effect standard score and a predetermined visual comfort standard score corresponding to each three-dimensional animation scene frame; obtaining the disparity map of each three-dimensional animation scene frame, extracting disparity statistic features of each three-dimensional animation scene frame based on its disparity map, and combining the disparity statistic features into one feature vector; and, determining the stereoscopic effect scoring model and the visual comfort scoring model for a three-dimensional animation scene frame respectively based on the feature vector of each three-dimensional animation scene frame in conjunction with the corresponding stereoscopic effect standard score and visual comfort standard score, in order to reduce the influence of subjective factors from the producers on the scoring.

First claim

What is claimed is:

1. A method performed by a device for determining scoring models of a three-dimensional animation scene frame, characterized in that the device comprising a processor, the processor executing the steps of:
obtaining a dataset of three-dimensional animation scene frames, the dataset comprising a first number of three-dimensional animation scene frames;
obtaining a predetermined stereoscopic effect standard score and a predetermined visual comfort standard score corresponding to each three-dimensional animation scene frame;
obtaining a disparity map of each three-dimensional animation scene frame, extracting disparity statistic features of each three-dimensional animation scene frame based on its disparity map, and combining the disparity statistic features into one feature vector;
establishing a first regression function, wherein the stereoscopic effect score of a three-dimensional animation scene frame is represented by the first regression function of the feature vector, solving the first regression function by using the corresponding stereoscopic effect standard scores as the output value of the first regression function and by introducing a kernel function using the method of penalty factor-support vector machine ϵ-SVR for regression, and determining a second regression function obtained by solving the first regression function as the stereoscopic effect scoring model of a three-dimensional animation scene frame; establishing a third regression function, wherein the visual comfort score of a three-dimensional animation scene frame is represented by the third regression function of the feature vector, solving the third regression function by using the corresponding visual comfort standard scores as the output value of the third regression function and by introducing a kernel function using the method of penalty factor-support vector machine ϵ-SVR for regression, and determining a fourth regression function obtained by solving the third regression function as the visual comfort scoring model of a three-dimensional animation scene frame;
the step of extracting disparity statistic features of each three-dimensional animation scene frame based on its disparity map executed by the processor comprises:
calculating the disparity statistic features according to the equations below:

⁢

mean
⁢

⁢
disparity
⁢

⁢
MD
⁢

:

⁢

⁢
MD

=

1
MN

⁢

∑

i
=
1

M

⁢

∑

j
=
1

N

⁢

⁢

D
⁡

(

i
,
j

)

⁢

median
⁢

⁢
disparity
⁢

⁢
MED
⁢

:

⁢

⁢
MED

=

median
⁢

⁢

(

∑

i
=
1

M

⁢

∑

j
=
1

N

⁢

⁢

D
⁡

(

i
,
j

)

)

⁢

maximum
⁢

⁢
positive
⁢

⁢
disparity
⁢

⁢
MPD
⁢

:

⁢

⁢
MPD

=

max
⁢

∑

i
=
1

M

⁢

∑

j
=
1

N

⁢

Assignees

• Univ Beijing Posts & Telecomm

Inventors

• Liu Liang
• Ma Huadong
• Wang Zeyu
• Lu Dawei

Classifications

• G06V10/764

  using classification, e.g. of video objects · CPC title

• H04N13/239 Primary

  using two two-dimensional [2D] image sensors having a relative position equal to or related to the interocular distance (H04N13/243 takes precedence) · CPC title

• G06T13/20 Primary

  Three-dimensional [3D] animation · CPC title

• H04N13/271

  wherein the generated image signals comprise depth maps or disparity maps · CPC title

• H04N2013/0081

  Depth or disparity estimation from stereoscopic image signals · CPC title