Image recognition method, image recognition apparatus, and recording medium

Patent metadata

Field	Value
Publication number	US-10438059-B2
Application number	US-201615222639-A
Country	US
Kind code	B2
Filing date	Jul 28, 2016
Priority date	Aug 5, 2015
Publication date	Oct 8, 2019
Grant date	Oct 8, 2019

Abstract

An image recognition method for an image recognition apparatus includes detecting, setting, acquiring, selecting, and specifying. At least one part of an identification target is selected from an identification target image. An inquiry region is set based on the detected part. A feature amount of the set inquiry region is acquired. At least one instance image corresponding to the identification target image is selected based on the acquired feature amount. A specific region of the identification target from the identification target image is specified based on the selected instance image.

First claim

What is claimed is:

1. An image recognition method, the image recognition method comprising:
detecting at least one part of an identification target from an identification target image;
setting an inquiry region based on the detected at least one part of the identification target from an identification target image;
acquiring a first feature amount of the set inquiry region;
selecting at least one first instance image corresponding to the identification target image from learning images based on the acquired first feature amount of the set inquiry region;
extracting a specific region of the identification target from the identification target image as a first extracted specific region based on the selected at least one first instance image;
acquiring a second feature amount of the first extracted specific region;
selecting at least one second instance image corresponding to the identification target image from the learning images based on the acquired second feature amount of the first extracted specific region;
generating a detector based on (i) the at least one first instance image selected based on the acquired first feature amount of the set inquiry region, and (ii) the at least one second instance image selected based on the acquired second feature amount of the first extracted specific region; and
extracting a specific region from the identification target image as a second extracted specific region using the generated detector,
wherein the identification target is a person, and the first extracted specific region is a first skin region of the person and the second extracted specific region is a second skin region of the person.

2. The image recognition method according to claim 1 , further comprising generating a dictionary based on the selected at least one first instance image,
wherein the second extracted specific region is extracted based on the generated dictionary.

3. The image recognition method according to claim 1 , wherein setting includes setting the inquiry region within a range where the identification target exists.

4. The image recognition method according to claim 1 , wherein setting includes setting the inquiry region based on a combination of more than one detected at least one part of the identification target from the identification target image.

5. The image recognition method according to claim 1 , further comprising acquiring a user setting result, wherein detecting includes detecting the at least one part of the identification target from the identification target image based on the user setting result.

6. The image recognition method according to claim 1 , further comprising using the second extracted specific region for post processing.

7. The image recognition method according to claim 1 , further comprising identifying the second skin region of the person by use of the at least one second instance image.

8. An image recognition method, the image recognition method comprising:
detecting at least one part of an identification target from an identification target image;
setting an inquiry region based on the detected at least one part of the identification target from the identification target image;
acquiring a first feature amount of the set inquiry region;
generating at least one first dictionary based on the acquired first feature amount of the set inquiry region;
extracting a specific region of the identification target from the identification target image as a first extracted specific region based on (i) the acquired first feature amount of the set inquiry region and (ii) the at least one first dictionary generated based on the acquired first feature amount of the set inquiry region;
acquiring a second feature amount of the first extracted specific region;
generating at least one second dictionary based on the acquired second feature amount of the first extracted specific region;
generating a detector based on (i) the at least one first dictionary generated based on the acquired first feature amount of the set inquiry region, and (ii) the at least one second dictionary generated based on the acquired second feature amount of the first extracted specific region; and
extracting a specific region from the identification target image as a second extracted specific region using the generated detector,
wherein the identification target is a person, and the first extracted specific region is a first skin region of the person and the second extracted specific region is a second skin region of the person.

9. The image recognition method according to claim 8 , wherein a plurality of dictionaries is weighed based on the acquired first feature amount, and generating the at least one first dictionary based on the acquired first feature amount includes selecting the weighed plurality of dictionaries.

10. An image recognition apparatus comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, causes the image recognition apparatus to perform operations including:
detecting at least one part of an identification target from an identification target image,
setting an inquiry region based on the detected at least one part of the identification target from the identification target image,
acquiring a first feature amount of the set inquiry region,
selecting at least one first instance image corresponding to the identification target image from learning images based on the acquired first feature amount of the set inquiry region,
extracting a specific region of the identification target from the identification target image as a first extracted specific region based on the selected at least one first instance image,
acquiring a second feature amount of the first extracted specific region,
selecting at least one second instance image corresponding to the identification target image from the learning images based on the acquired second feature amount of the first extracted specific region,
generating a detector based on (i) the at least one first instance image selected based on the acquired first feature amount of the set inquiry region, and (ii) the at least one second instance image selected based on the acquired second feature amount of the first extracted specific region, and
extracting a specific region from the identification target image as a second extracted specific region using the generated detector,
wherein the identification target is a person, and the first extracted specific region is a first skin region of the person and the second extracted specific region is a second skin region of the person.

11. An image recognition apparatus comprising:
one or more processors; and
memory storing instructions that, when executed by the one or more processors, causes the image recognition apparatus to perform operations including:
detecting at least one part of an identification target from an identification target image,
setting an inquiry region based on the detected at least one part of the identification target from the identification target image,
acquiring a first feature amount of the set inquiry region,
generating at least one first dictionary based on the acquired first feature amount of the set inquiry region,
extracting a specific region of the identification target from the identification target image as a first extracted specific region based on (i) the acquired first feature amount of the set inquiry region and (ii) the at least one first dictionary generated based on the acquired first feature amount of the set inquiry region,
acquiring a second feature amount of the first extracted specific region,
generating at least one second dictionary based on the a

Assignees

• Canon Kk

Inventors

• Saruta Takayuki
• Tate Shunta

Classifications

• G06V40/161 Primary

  Detection; Localisation; Normalisation · CPC title

• G06V40/10

  Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands · CPC title

• G06K2009/4666

  Physics · mapped topic

• G06K9/00228

  Physics · mapped topic

• G06K9/00362 Primary

  Physics · mapped topic