US 11,694,315 B2Granted

Artificial intelligence software for document quality inspection

AssigneeKYOCERA Document Solutions Inc.

InventorsKilho Shin, Kendrick Esperanza Wong

Life of this patent

Estimating the term…

Loading patent…

Abstract

A system employs a trained model to detect artifact(s) associated with artifact type(s) appearing in a reproduction of a source image (a test image). The system determines differences between the test image and the source image and outputs probabilities that the artifact(s) in the test image are associated with each of the artifact type(s). A dataset for training the model includes: (i) a reference category including reference image(s) without any artifacts; and (ii) artifact categories, each corresponding to a respective one of the artifact types and including noised images associated with the respective artifact type. Each noised image includes one of the reference images and an artifact associated with the respective artifact type. The model is trained to detect the artifact type(s) by providing the model with the dataset and causing the model to process differences between each noised image and the reference image in the noised image.

Claims20

Dependent claims

A method for producing a system to determine quality of images produced by a printer, the method comprising:

determining one or more artifact types, wherein one or more artifacts associated with the one or more artifact types appear in test images produced by one or more printers, the test images being reproductions of source images;
generating a dataset including categories of images, the categories of images including:
a reference category including one or more reference images without any artifacts associated with the one or more artifact types, and
one or more artifact categories, each artifact category corresponding to a respective one of the artifact types and including noised images associated with the respective artifact type, each noised image including one of the reference images and an artifact associated with the respective artifact type;
training a model implemented on a computer system, wherein the training includes:
providing the model with the dataset;
and
causing the model to process differences between each noised image and the reference image included in the noised image thereby allowing the model to detect the one or more artifacts associated with the one or more artifact types;
and
configuring the model to:
receive the test images produced by the one or more printers and the source images corresponding to the test images;
detect the one or more artifacts associated with the one or more artifact types appearing in the test images relative to the corresponding source images reproduced in the test images;
and
output information relating to the one or more artifacts detected by the model in the test images.

11 more elements

A system for determining quality of images produced by a printer, the system comprising:

one or more computer storage devices configured to store a model to detect one or more artifacts associated with one or more artifact types appearing in a test image produced by a printer, the test image being a reproduction of a source image;
and
one or more processors configured to execute instructions implementing the model, the instructions causing the one or more processors to:
receive the test image and the source image;
determine differences between the test image and the source image;
calculate, based on the differences between the test image and the source image, probabilities that the one or more artifacts in the test image are associated with each of the one or more artifact types;
and
output the probabilities to indicate the quality of the test image relative to the source image,
wherein the model is trained to detect the one or more artifact types by:
generating a dataset including categories of images, the categories of images including:
a reference category including one or more reference images without any artifacts associated with the one or more artifact types, and
one or more artifact categories, each artifact category corresponding to a respective one of the artifact types and including noised images associated with the respective artifact type, each noised image including one of the reference images and an artifact associated with the respective artifact type;
providing the model with the dataset;
and
causing the model to process differences between each noised image and the reference image included in the noised image.

12 more elements

One or more non-transitory computer-readable storage media, having computer-executable instructions stored thereon, wherein when executed by a one or more processors, the computer-executable instructions cause the one or more processors to:

receive a test image produced by a printer and a source image, wherein the test image is a reproduction of the source image, and one or more artifacts associated with one or more artifact types appear in the test image;
determine differences between the test image and the source image;
calculate, based on the differences between the test image and the source image, probabilities that the one or more artifacts in the test image are associated with each of the one or more artifact types;
and
output the probabilities to indicate a quality of the test image relative to the source image,
wherein the instructions implement a model that is trained to detect the one or more artifact types by:
generating a dataset including categories of images, the categories of images including:
a reference category including one or more reference images without any artifacts associated with the one or more artifact types, and
one or more artifact categories, each artifact category corresponding to a respective one of the artifact types and including noised images associated with the respective artifact type, each noised image including one of the reference images and an artifact associated with the respective artifact type;
providing the model with the dataset;
and
causing the model to process differences between each noised image and the reference image included in the noised image.

9 more elements

Dependencies are inferred from the claim text.

Facts

Application: 17/243,887Filed 29 Apr 2021
Published: US 2022/0351356 A13 Nov 2022
Granted: 4 Jul 2023
Term ends: Not estimatedThe source records don’t support an estimate.

Classification16 codes

G06N 3/084
Learning methods: backpropagation, e.g. using gradient descent
G06T 7/0002
Image analysis: inspection of images, e.g. flaw detection
G06N 3/045
Architecture, e.g. interconnection topology: combinations of networks
G06N 3/0464
Architecture, e.g. interconnection topology: convolutional networks [CNN, ConvNet]