[Retracted] Local Privacy Protection for Sensitive Areas in Multiface Images
Algorithm 3
PPSA.
Input: Original image X, privacy budget , estimated number of faces , preset parameters , estimate the number of faces in the image , subgraph similarity expectation
Output: Image satisfying differential privacy.
(1)
Read the original image X, convert the image into gray matrix and store it in matrix
(2)
MTCNN is used to extract the face region and face feature point in face image X
(3)
Set the gray matrix X according to the preset parameters β It is divided into subgraph sets with the same structure
(4)
If belongs to a subgraph , set as
(5)
Find all backup
(6)
Adjust the face image area as according to the sub image size
(7)
(8)
(9)
(10)
Set used to record whether this seed has been used during regional growth
(11)
Create a linked list to record the status of the current seed region merging process
(12)
)
(13)
(14)
(15)
Pick unmarked adds noise to it as ,
(16)
(17)
Attempt to merge adjacent start regional growth
(18)
(19)
brake
(20)
else
(21)
(22)
(23)
else
(24)
(25)
(26)
is set to marked,
(27)
Release all data in the list, and the consolidation of this subregion is completed
(28)
else
(29)
Execute line 15–line 28, line 16 adjusted to , line 17 is adjusted to
