A two-stage secure multi-party computation method for image and text localization and recognition
A technology for secure multi-party computing and text positioning, applied in computing, computer components, neural learning methods, etc., can solve problems such as the inability to meet privacy protection requirements, and achieve high security and reliable services.
Active Publication Date: 2022-07-15
SHANGHAI JIAOTONG UNIV
View PDF12 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0002] At present, most cloud services that provide image text positioning and recognition algorithms require real image information. When users accept such services, they often need to provide real images to the server, which cannot meet the needs of privacy protection.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0025] like figure 1As shown, the present embodiment relates to a method for locating and recognizing image text based on the privacy protection of secure multi-party computing, wherein the involved server provides a remote calling interface based on services such as HTTP, and the models for text locating and recognition are trained in plaintext respectively. Deployed on the server side, the method specifically includes the following steps:
[0026] 1) Based on the function secret sharing protocol in the secure multi-party computing, the user encrypts the information of each pixel in the picture through the secret function therein, and transmits the encrypted picture information to the cloud server;
[0027] 2) The cloud server performs feature extraction on the image information based on the trained single-character text positioning model, and obtains three-level encrypted image single-character positioning feature map F char1 , F char2 and F char3 , and transmit the encry...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
A two-stage secure multi-party computing privacy-preserving image text location and recognition method. The text location and recognition scheme for pictures does not involve plaintext transmission, and the image information and the transmission content between the client and the client are encrypted by secure multi-party computing. Yes, it meets the requirements of image privacy protection, the service is reliable, and the security is high. The invention separates encrypted character location and recognition network services, and proposes segmentation-based single-character location and full-convolution single-character recognition network suitable for a secure multi-party computing framework, which can respectively provide encrypted word location and single-character recognition services.
Description
technical field [0001] The invention relates to a technology in the field of image processing, in particular to a privacy-preserving image and text location and recognition method based on two-stage secure multi-party computation, in particular to the realization of using secure multi-party computation technology without revealing real picture information A method for word location and recognition of pictures. Background technique [0002] At present, most cloud services that provide image text positioning and recognition algorithms require real image information. When users accept such services, they often need to provide real images to the server, which cannot meet the needs of privacy protection. The privacy computing technology represented by secure multi-party computing realizes multi-party computing tasks on the basis of protecting the data itself from external leakage, solves the contradiction between availability and privacy in data circulation, pays attention to the...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06V30/14G06V30/148G06V30/19G06K9/62G06F21/60G06F21/62G06N3/04G06N3/08
CPCG06F21/602G06F21/6245G06N3/08G06V10/22G06V30/153G06N3/045G06F18/213G06F18/24G06F18/214
Inventor 茹超飞黄征郭捷邱卫东
Owner SHANGHAI JIAOTONG UNIV