# US 12,183,101 B2 · Image processing system, image processing method, and storage medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,183,101 B2
- Document ID: US12183101B2
- Status: Granted
- Kind: B2
- Application number: 17/716,084
- Filed: 2022-04-08
- Published: 2022-10-20
- Granted: 2024-12-31
- Earliest claimed: 2021-04-12
- Assignees: Canon Kabushiki Kaisha
- Applicants: Canon Kabushiki Kaisha
- Inventors: Yuki Narumi; Go Oishi
- Web page: https://patentgrep.com/patent/US12183101B2

## CPC

- G06V 30/1463: {Orientation detection or correction, e.g. rotation of multiples of 90 degrees}
- G06V 30/226: of cursive writing
- G06V 10/82: using neural networks
- G06V 30/1475: {Inclination or skew detection or correction of characters or of image to be recognised}
- G06V 30/148: Segmentation of character regions
- G06V 30/1801: {Detecting partial patterns, e.g. edges or contours, or configurations, e.g. loops, corners, strokes or intersections (extracting features by contour coding G06V30/182)}
- G06V 30/19147: {Obtaining sets of training patterns; Bootstrap methods, e.g. bagging or boosting}
- G06V 30/22: characterised by the type of writing
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

An image processing system performs tilt correction with respect to a document image having handwritten characters and typed letters mixed with each other. The image processing system separates the document image into an image with handwritten characters determined as handwritten characters and an image without handwritten characters not determined as handwritten characters, estimates a tilt angle of the image without handwritten characters, and corrects the document image on the basis of the tilt angle.

## Claims

### Claim 1

An image processing system which performs tilt correction with respect to a document image, the image processing system comprising:
at least one memory that stores instructions; and
at least one processor that executes the instructions to perform:
(1) obtaining the document image, wherein the document image includes handwritten characters and typed letters;
(2) generating a first image containing no handwritten characters from the document image, wherein (a) in a case where a range of a tilt angle of the document image is determined to be equal to or larger than a prescribed value, the first image containing no handwritten characters is generated by determining pixels of handwritten characters using a first neural network and by eliminating the determined pixels of handwritten characters from the document image, (b) in a case where the range of the tilt angle of the document image is determined to be smaller than the prescribed value, the first image containing no handwritten characters is generated by determining the pixels of handwritten characters using a second neural network and by eliminating the determined pixels of handwritten characters from the document image, and (c) the second neural network is different from the first neural network;
(3) estimating a tilt angle of the generated first image; and
(4) tilt-correcting the document image on the basis of the estimated tilt angle of the generated first image,
wherein the tilt-corrected document image includes the handwritten characters and the typed letters.

### Claim 2 (depends on 1)

The image processing system according to claim 1, wherein a tilt of a document image used in learning in the second neural network is smaller than a tilt of a document image used in learning in the first neural network.

### Claim 3 (depends on 1)

The image processing system according to claim 1, wherein the first neural network and the second neural network is are obtained by learning a plurality of images of handwritten characters having different tilt angles.

### Claim 4 (depends on 1)

The image processing system according to claim 1, wherein the first neural network and the second neural network is are obtained by learning a plurality of document images having different tilt angles.

### Claim 5 (depends on 1)

The image processing system according to claim 1, wherein a network structure of the second neural network is simpler than a network structure of the first neural network.

### Claim 6

An image processing method for performing tilt correction with respect to a document image, the image processing method comprising:
obtaining the document image, wherein the document image includes handwritten characters and typed letters;
generating a first image containing no handwritten characters from the document image, wherein (a) in a case where a range of a tilt angle of the document image is determined to be equal to or larger than a prescribed value, the first image containing no handwritten characters is generated by determining pixels of handwritten characters using a first neural network and by eliminating the determined pixels of handwritten characters from the document image, (b) in a case where the range of the tilt angle of the document image is determined to be smaller than the prescribed value, the first image containing no handwritten characters is generated by determining the pixels of handwritten characters using a second neural network and by eliminating the determined pixels of handwritten characters from the document image, and (c) the second neural network is different from the first neural network;
estimating a tilt angle of the generated first image; and
tilt-correcting the document image on the basis of the estimated tilt angle of the generated first image,
wherein the tilt-corrected document image includes the handwritten characters and the typed letters.

### Claim 7

A non-transitory computer-readable storage medium storing a program causing a computer to execute a method comprising:
obtaining a document image, wherein the document image includes handwritten characters and typed letters;
generating a first image containing no handwritten characters from the document image, wherein (a) in a case where a range of a tilt angle of the document image is determined to be equal to or larger than a prescribed value, the first image containing no handwritten characters is generated by determining pixels of handwritten characters using a first neural network and by eliminating the determined pixels of handwritten characters from the document image, (b) in a case where the range of the tilt angle of the document image is determined to be smaller than the prescribed value, the first image containing no handwritten characters is generated by determining the pixels of handwritten characters using a second neural network and by eliminating the determined pixels of handwritten characters from the document image, and (c) the second neural network is different from the first neural network;
estimating a tilt angle of the generated first image; and
tilt-correcting the document image on the basis of the estimated tilt angle of the generated first image,
wherein the tilt-corrected document image includes the handwritten characters and the typed letters.