Back-propagation image visual saliency detection method based on depth image mining

Patent metadata

Field	Value
Publication number	US-11227178-B2
Application number	US-201716336737-A
Country	US
Kind code	B2
Filing date	Nov 24, 2017
Priority date	Jun 29, 2017
Publication date	Jan 18, 2022
Grant date	Jan 18, 2022

Abstract

A back-propagation significance detection method based on depth map mining, comprising: for an input image Io, at a preprocessing phase, obtaining a depth image Id and an image Cb with four background corners removed of the image Io; at a first processing phase, carrying out positioning detection on a significant region of the image by means of the obtained image Cb with four background corners removed and the obtained depth image Id to obtain the preliminary detection result S1 of a significant object in the image; then carrying out depth mining on a plurality of processing phases of the depth image Id to obtain corresponding significance detection results; and then optimizing the significance detection result mined in each processing phase by means of a back-propagation mechanism to obtain a final significance detection result map. The method can improve the detection accuracy of the significance object.

First claim

The invention claimed is:

1. A back-propagation saliency detection method based on depth image mining, comprising, for an input image I o :
at a preprocessing phase, obtaining a depth image I d of an image I o and an image C b with four background corners removed;
at a first processing phase, carrying out positioning detection on a salient region of the image by means of the obtained image C b with four background corners removed and the depth image I d to obtain a preliminary detection result of a salient object in the image, comprising steps 11-14;
Step 11, dividing the image into K regions by means of a K-means algorithm, and calculating a color saliency value S c (r k ) of each subregion through the formula (1):

S
c

⁡

(

r
k

)

=

∑

i
=
1

,

i
≠
k

K

⁢

⁢

P
i

⁢

W
s

⁡

(

r
k

)

⁢

D
c

⁡

(

r
k

,

r
i

)

(
1
)

wherein r k and r i respectively represent regions k and i, D c (r k , r i ) represents a Euclidean distance of the region k and the region i in an L*a*b color space, P i represents a proportion of the region i to an image region; W s (r k ) is obtained through the formula (2):

W
s

⁡

(

r
k

)

=

e

-

D
o

⁡

(

r
k

,

r
i

)

σ
2

(
2
)

wherein D o (r k , r i ) represents a coordinate position distance of the region k and the region i, and σ is a parameter controlling the range of the W s (r k );
Step 12, calculating a depth saliency value S d (r k ) of the depth image through the formula (3):

S
d

⁡

(

r
k

)

=

∑

i
=
1

,

i
≠
k

K

⁢

P
i

⁢

W
s

⁡

(

r
k

)

⁢

D
d

⁡

(

r
k

Assignees

• Univ Peking Shenzhen Graduate School

Inventors

• Li Ge
• Zhu Chunbiao
• Wang Wenmin
• Wang Ronggang
• Huang Tiejun
• Gao Wen

Classifications

• G06T7/73 Primary

  using feature-based methods · CPC title

• G06V10/763

  Non-hierarchical techniques, e.g. based on statistics of modelling distributions · CPC title

• G06V10/462

  Salient features, e.g. scale invariant feature transforms [SIFT] · CPC title

• G06F18/23213

  with fixed number of clusters, e.g. K-means clustering · CPC title

• G06V10/56

  relating to colour · CPC title