US 11,783,625 B2
Method for verifying the identity of a user by identifying an object within an image that has a biometric characteristic of the user and separating a portion of the image comprising the biometric characteristic from other portions of the image
Hardik Gupta, Jabalpur (IN); Satheesh Murugan, Ramapurum (IN); and Jesus Aragon, Majadahonda (ES)
Assigned to Identy Inc., Dover, DE (US)
Appl. No. 17/433,318
Filed by Identy Inc., Dover, DE (US)
PCT Filed Feb. 24, 2020, PCT No. PCT/IB2020/051523
§ 371(c)(1), (2) Date Aug. 24, 2021,
PCT Pub. No. WO2020/174350, PCT Pub. Date Sep. 3, 2020.
Claims priority of application No. 19382137 (EP), filed on Feb. 26, 2019.
Prior Publication US 2022/0157077 A1, May 19, 2022
Int. Cl. G06K 9/20 (2006.01); G06K 9/62 (2022.01); G06T 5/00 (2006.01); G06V 40/12 (2022.01); G06V 10/25 (2022.01); G06V 40/13 (2022.01); G06N 3/08 (2023.01)
CPC G06V 40/1365 (2022.01) [G06N 3/08 (2013.01); G06V 10/25 (2022.01); G06V 40/1318 (2022.01)] 17 Claims
OG exemplary drawing
 
1. A method comprising:
obtaining, by a processor from an optical sensor of a mobile device, an image of an object;
processing, by the processor, the image using a neural network, wherein processing the image comprises distinguishing, using the neural network, a first portion of the image comprising a region of interest, ROI, from a second portion of the image;
after processing the image, extracting, by the processor, the ROI from the image;
extracting, by the processor, a biometric characteristic of the object from the ROI; and
processing, by the processor, the biometric characteristic of the object to determine whether the biometric characteristic of the object identifies a user, wherein processing the image comprises processing the image by an encoder to obtain an encoded image and processing the encoded image by a decoder to obtain a decoded output image, wherein processing the image comprises distinguishing a first portion of the decoded output image from a second portion of the decoded output image, wherein, in the decoded image, each pixel has a value x∈]−∞; +∞[ and, before distinguishing of the first portion of the decoded image from the second portion of the decoded image, an activation function

OG Complex Work Unit Math
is applied to each pixel in the decoded image and a normalized value x=S(x) is associated with each pixel in the decoded image, wherein the distinguishing is performed on the pixels with the values x.