Automatic orientation correction for captured images

Patent metadata

Field	Value
Publication number	US-12561934-B2
Application number	US-202318182536-A
Country	US
Kind code	B2
Filing date	Mar 13, 2023
Priority date	Mar 13, 2023
Publication date	Feb 24, 2026
Grant date	Feb 24, 2026

Abstract

In some implementations, a device may receive an image of a document, the image depicting a reference feature associated with the document, the reference feature including at least one of: a face of a person, a machine-readable code, or a text field. The device may identify a rotational angle of the reference feature as depicted in the image based on comparing the reference feature as depicted in the image to one or more orientation parameters of the reference feature associated with a display orientation associated with the document. The device may rotate the image of the document by an angle to obtain an orientated image of the document, the angle being based on the rotational angle of the reference feature as depicted in the image. The device may provide the orientated image of the document for display.

First claim

What is claimed is:

1 . A system for automatic orientation correction for captured images, the system comprising:
one or more memories; and one or more processors, communicatively coupled to the one or more memories, configured to:
obtain, via a user device, an image of an identification document, the identification document including a facial image depicting a face of a person associated with the identification document;
select a machine learning model, from a plurality of machine learning models, based on at least one of one or more reference features associated with the identification document or a document type associated with the identification document,
wherein the machine learning model is a facial recognition model that is trained to identify rotational angles of faces, and
wherein one of the one or more reference features is associated with a facial image;

input, to the selected machine learning model, the image of the identification document;
obtain, from the selected machine learning model, an indication of a rotational angle of the face depicted in the facial image based on inputting, to the selected machine learning model, the image of the identification document;
rotate the image of the identification document by an angle to obtain an orientated image of the identification document, the angle being based on the rotational angle of the face and a reference angle associated with a display orientation of the identification document;
provide the orientated image of the identification document for display;
receive feedback information indicating whether the oriented image has been accepted;
transmit, based on the feedback information, the oriented image or the image of the identification document to a server device for use in an authentication procedure related to the user device; and
re-train, based on the feedback information, the selected machine learning model.

2 . The system of claim 1 , wherein the one or more processors are further configured to:
modify the display orientation based on the feedback information indicating that the orientated image of the identification document has been rejected.

3 . The system of claim 1 , wherein the display orientation of the identification document is an orientation at which images of the identification document are to be viewed.

4 . The system of claim 1 , wherein the one or more processors are further configured to:
analyze, using an optical character recognition (OCR) technique, one or more text fields as depicted in the orientated image of the identification document; and detect whether the orientated image of the identification document is in a correct orientation based on whether text included in the one or more text fields is detectable using the OCR technique.

5 . The system of claim 1 , wherein the one or more processors are further configured to:
identify an aspect ratio for a review page associated with the image of the identification document based on the orientated image of the identification document.

6 . The system of claim 5 , wherein the one or more processors, to provide the orientated image of the identification document for display, are configured to:
provide the orientated image of the identification document for display in the review page having the aspect ratio.

7 . The system of claim 1 , wherein the one or more processors, when re-training the selected machine learning model, are configured to:
re-train the selected machine learning model with the feedback information to identify the rotational angles of faces.

8 . The system of claim 7 , wherein re-training the selected machine learning model with the feedback information improves an accuracy of rotational angle determinations.

9 . A method of automatic orientation correction for captured images, comprising:
receiving, by a device, an image of a document, the image depicting a reference feature associated with the document, the reference feature including at least one of:
a face of a person,
a machine-readable code, or
a text field;
selecting, by the device, a machine learning model, from a plurality of machine learning models, based at least on one of the reference feature or a document type associated with the document; identifying, by the device and using the selected machine learning model, a rotational angle of the reference feature as depicted in the image based on comparing the reference feature as depicted in the image to one or more orientation parameters of the reference feature associated with a display orientation associated with the document; rotating, by the device, the image of the document by an angle to obtain an orientated image of the document, the angle being based on the rotational angle of the reference feature as depicted in the image; providing, by the device, the orientated image of the document for display; receiving, by the device, feedback information indicating whether the oriented image has been accepted; transmitting, by the device and based on the feedback information, the oriented image or the image of the document to a server device as part of an authentication procedure; and re-training, by the device and based on the feedback information, the selected machine learning model.

10 . The method of claim 9 , wherein identifying the rotational angle of the reference feature comprises:
inputting, to the selected machine learning model, the image of the document, the selected machine learning model being trained to identify rotational angles of the reference feature; and obtaining, from the selected machine learning model, an indication of the rotational angle of the reference feature as depicted in the image based on inputting the image of the document to the machine learning model.

11 . The method of claim 9 , wherein identifying the rotational angle of the reference feature comprises:
identifying a first reference point and a second reference point included in the reference feature; and determining the rotational angle based on a difference between a first angle between the first reference point and the second reference point as depicted in the image and a second angle associated with a known orientation of the first reference point and the second reference point.

12 . The method of claim 9 , wherein the reference feature includes the face of the person, the method further comprising:
analyzing, using a computer vision technique, the face of the person as depicted in the orientated image to identify one or more facial features of the face; and determining whether the orientated image is in a correct orientation based on whether the one or more facial features of the face are in known facial positions associated with respective facial features from the one or more facial features.

13 . The method of claim 9 , wherein the reference feature includes the machine-readable code, the method further comprising:
detecting one or more patterns included in the machine-readable code as depicted in the orientated image; and determining whether the orientated image is in a correct orientation based on whether the one or more patterns are in an orientation associated with the machine-readable code.

14 . The method of claim 9 , wherein re-training the selected machine learning model comprises:
re-training the selected machine learning model with the feedback information to identify rotational angles of faces.

15 . A non-transitory computer-readable medium storing a set of instructions, the set of instructions comprising:
one or more instructions that, when executed by one or more proc

Assignees

• Capital One Services Llc

Inventors

• Pribble Jason
• Patil Swapnil
• Tran Timothy

Classifications

• G06V40/161

  Detection; Localisation; Normalisation · CPC title

• G06V10/70

  using pattern recognition or machine learning (optical pattern recognition or electronic computations therefor G06V10/88) · CPC title

• G06V40/171

  Local features and components; Facial parts (eye characteristics G06V40/18); Occluding parts, e.g. glasses; Geometrical relationships · CPC title

• G06V30/10

  Character recognition · CPC title

• B42D25/23

  Identity cards · CPC title