Method for correcting image by device and device therefor

Patent metadata

Field	Value
Publication number	US-2019279345-A1
Application number	US-201716347330-A
Country	US
Kind code	A1
Filing date	Nov 8, 2017
Priority date	Nov 8, 2016
Publication date	Sep 12, 2019
Grant date	—

Abstract

The device may identify a plurality of objects and use a rule-based or artificial intelligence (AI) algorithm when determining a plurality of correction filters respectively corresponding to the plurality of identified objects. When identifying the plurality of objects using the AI algorithm and determining the plurality of correction filters respectively corresponding to the plurality of identified objects, the device may use machine learning, a neural network, or a deep learning algorithm.

First claim

1 . A device for correcting an image, the device comprising:
a memory storing one or more instructions; and a processor configured to execute the one or more instructions stored in the memory, wherein the processor, by executing the one or more instructions, is further configured to obtain an image comprising a plurality of objects, identify the plurality of objects in the image based on a result of using one or more neural networks, determine a plurality of correction filters respectively corresponding to the plurality of identified objects, and correct the plurality of objects in the image, respectively, by using the plurality of determined correction filters.

2 . The device of claim 1 , wherein the processor, by executing the one or more instructions, is further configured to identify the plurality of objects and determine the plurality of correction filters, based on a training result of using a predetermined plurality of image attributes used for identifying the plurality of objects and a predetermined plurality of display attributes of objects used for determining the plurality of correction filters.

3 . The device of claim 1 , wherein the processor, by executing the one or more instructions, is further configured to learn a criterion for identifying the plurality of objects and determining the plurality of correction filters, by using a first network for determining a category corresponding to each of the plurality of objects in the neural network and a second network for determining the plurality of correction filters respectively corresponding to the plurality of objects.

4 . The device of claim 2 , further comprising: a user interface configured to receive a user input,
wherein the processor, by executing the one or more instructions, is further configured to control a display to display the image on a screen of the device, control the user interface to receive the user input that touches one of the plurality of objects in the displayed image, and obtain a correction filter corresponding to an object selected from the plurality of objects when location information touched by the user input is input to the neural network.

5 . The device of claim 1 , wherein the processor, by executing the one or more instructions, is further configured to obtain history information indicating the plurality of correction filters set by a user with respect to the plurality of identified objects, respectively, before correction of the image, and, based on the obtained history information, determine the plurality of correction filters respectively corresponding to the plurality of identified objects.

6 . The device of claim 1 , wherein the processor, by executing the one or more instructions, is further configured to select an object related to a user, from among the plurality of identified objects, based on user information previously stored in the device, determine a correction filter corresponding to the selected object, and correct the selected object by using the determined correction filter.

7 . The device of claim 1 , wherein the processor is further configured to identify a subject of content comprising the image and determine the plurality of correction filters respectively corresponding to the plurality of identified objects based on the identified subject of the content and the plurality of identified objects.

8 . A method of correcting an image, the method comprising:
obtaining an image comprising a plurality of objects; identifying the plurality of objects in the image; determining a plurality of correction filters respectively corresponding to the plurality of identified objects; and correcting the plurality of objects in the image, respectively, by using the plurality of determined correction filters, wherein the identifying of the plurality of objects and the determining of the plurality of correction filters comprise identifying the plurality of objects and determining the plurality of correction filters based on a result of using one or more neural networks.

9 . The method of claim 8 , wherein the identifying of the plurality of objects and the determining of the plurality of correction filters comprise: identifying the plurality of objects and determining the plurality of correction filters, based on a training result of using a predetermined plurality of image attributes used for identifying the plurality of objects and a predetermined plurality of display attributes of objects used for determining the plurality of correction filters.

10 . The method of claim 8 , further comprising learning a criterion for identifying the plurality of objects and determining the plurality of correction filters, by using a first network for determining a category corresponding to each of the plurality of objects in the neural network and a second network for determining the plurality of correction filters respectively corresponding to the plurality of objects.

11 . The method of claim 9 , further comprising:
displaying the image on a screen of a device and receiving a user input that touches one of the plurality of objects in the displayed image, and wherein an object selected from the plurality of objects is identified and a correction filter corresponding to the selected object is determined when location information touched by the user input is input to the neural network.

12 . The method of claim 8 , further comprising: obtaining history information indicating the plurality of correction filters set by a user with respect to the plurality of identified objects, respectively, before correction of the image,
wherein the determining of the plurality of correction filters comprises; based on the obtained history information, determining the plurality of correction filters respectively corresponding to the plurality of identified objects.

13 . The method of claim 8 , further comprising: selecting an object related to a user among the plurality of identified objects based on user information previously stored in the device,
wherein the correcting comprises correcting the selected object by using the correction filter corresponding to the selected object.

14 . The method of claim 8 , further comprising identifying a subject of content comprising the image,
wherein the determining of the plurality of correction filters comprises; determining the plurality of correction filters respectively corresponding to the plurality of identified objects based on the identified subject of the content and the plurality of identified objects.

15 . A computer program product comprising a recording medium, the recording medium storing instructions for:
obtaining an image comprising a plurality of objects; identifying the plurality of objects in the image; determining a plurality of correction filters respectively corresponding to the plurality of identified objects; and correcting the plurality of objects in the image, respectively, by using the plurality of determined correction filters, wherein the identifying of the plurality of objects and the determining of the plurality of correction filters comprise identifying the plurality of objects and determining the plurality of correction filters based on a result of using one or more neural networks.

Assignees

• Samsung Electronics Co Ltd

Inventors

• Kim Sung-Hoon
• Kang Seong-Min

Classifications

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title

• G06T2200/24

  involving graphical user interfaces [GUIs] · CPC title

• G06F3/0488

  using a touch-screen or digitiser, e.g. input of commands through traced gestures · CPC title