Image generation method and apparatus

Patent metadata

Field	Value
Publication number	US-12555276-B2
Application number	US-202118245081-A
Country	US
Kind code	B2
Filing date	May 24, 2021
Priority date	Sep 14, 2020
Publication date	Feb 17, 2026
Grant date	Feb 17, 2026

Abstract

An image generation method and apparatus. The method includes: performing image composition on the basis of an acquired image generation condition, and determining a primitive type and layout information of an image, so as to obtain a composition image ( 201 ); according to scenario construction information corresponding to the image generation condition, performing scenario construction on the composition image, and determining a primitive corresponding to each primitive type from among a plurality of primitive types in the composition image, so as to obtain a scenario construction image ( 202 ); adjusting the color of each primitive in the scenario construction image according to color matching information corresponding to the image generation condition, so as to obtain a color-matched image ( 203 ); and rendering the color-matched image to obtain a target image ( 204 ).

First claim

What is claimed is:

1 . A method for generating an image, comprising:
performing an image composition based on an acquired image generation condition, and ascertaining a primitive type and layout information of an image to obtain a composition image, comprising performing, based on the acquired image generation condition, a multi-layer image composition from part to whole using a preset composition knowledge graph and ascertaining a primitive type and layout information in each layer of image to obtain a plurality of layers of composition images nested layer by layer and representing composition information of the layers, wherein the nested layer by layer is used to represent that, for each layer of composition image in the plurality of layers of composition images, the layer of composition image is nested with a lower layer of composition image that takes a part of primitive types in the layer of composition image as a whole; performing a scene construction on the composition image according to scene construction information corresponding to the acquired image generation condition and ascertaining a primitive corresponding to each primitive type in a plurality of primitive types in the composition image to obtain a scene construction image; adjusting a color of each primitive in the scene construction image according to color matching information corresponding to the acquired image generation condition to obtain a color matching image; and rendering the color matching image to obtain a target image.

2 . The method according to claim 1 , wherein performing, based on the acquired image generation condition, the multi-layer image composition from part to whole using the preset composition knowledge graph and ascertaining the primitive type and layout information in each layer of the image to obtain the plurality of layers of composition images nested layer by layer and representing composition information of the layers comprises:
performing, for the each layer of composition image in the plurality of layers of composition images, following operations:
ascertaining, for each composition parameter in a plurality of composition parameters ascertained based on the acquired image generation condition and the preset composition knowledge graph, a parameter value of the composition parameter through determination information corresponding to the composition parameter in a composition process of the layer of composition image, wherein the composition parameter comprises an operation parameter for each primitive type in the layer of composition image and a call parameter for the lower layer of composition image; and
obtaining the layer of composition image based on each parameter value.

3 . The method according to claim 2 , wherein, for the layer of composition image, parameter values ascertained by a given composition parameter are different, and the layer of composition image comprises a plurality of sub-composition images having identical primitive types and different layouts, and
obtaining the layer of composition image based on each parameter value comprises:
performing, for each composition parameter in the plurality of composition parameters in the composition process of the layer of composition image, a nested call on a parameter value of a composition parameter having a previously ascertained parameter value, to ascertain a plurality of sets of parameter value groups comprising parameter values of the plurality of composition parameters; and
obtaining the plurality of sub-composition images having the identical primitive types and the different layouts and corresponding to the layer of composition image, based on the plurality of sets of parameter value groups.

4 . The method according to claim 3 , further comprising:
ascertaining, in response to ascertaining that a selection instruction is received, at least one sub-composition image of the layer of composition image on which an upper layer of composition image of the layer of composition image is to perform a nested call, based on the selection instruction.

5 . The method according to claim 1 , wherein performing the scene construction on the composition image according to the scene construction information corresponding to the acquired image generation condition and ascertaining a primitive corresponding to each primitive type in the plurality of primitive types in the composition image to obtain the scene construction image comprises:
performing, for each layer of composition image in the plurality of layers of composition images, following operations:
performing a first knowledge transfer operation, until it is ascertained that a first preset termination condition is reached;
ascertaining final scene construction information corresponding to the layer of composition image according to target scene construction information obtained through the first knowledge transfer operation for each time; and
performing a scene construction on the layer of composition image according to the final scene construction information, and ascertaining a primitive corresponding to each primitive type in a plurality of primitive types in the layer of composition image to obtain a scene construction image corresponding to the layer of composition image,
wherein the first knowledge transfer operation comprises: ascertaining a first target primitive type set corresponding to a first primitive type set based on a similarity; ascertaining scene construction information corresponding to the first target primitive type set as the target scene construction information; and ascertaining a primitive type included in the first primitive type set but not included in the first target primitive type set as a first primitive type set for the first knowledge transfer operation for next time, wherein a first primitive type set for the first knowledge transfer operation for first time is a primitive type set corresponding to the layer of composition image.

6 . The method according to claim 5 , wherein ascertaining the primitive corresponding to each primitive type in the plurality of primitive types in the layer of composition image to obtain the scene construction image corresponding to the layer of composition image comprises:
ascertaining, from a primitive library, the primitive corresponding to the each primitive type in the plurality of primitive types in the layer of composition image according to a pre-established search tree representing a relationship between a primitive type and a primitive, to obtain the scene construction image corresponding to the layer of composition image.

7 . The method according to claim 5 , wherein the first preset termination condition comprises: the first primitive type set being an empty set, and a first primitive type set for the first knowledge transfer operation for this time being identical to a first primitive type set for the first knowledge transfer operation for last time, and
the method further comprises:
adding, in response to ascertaining that the first preset termination condition is that the first primitive type set for the first knowledge transfer operation for this time is identical to the first primitive type set for the first knowledge transfer operation for last time, information representing a scene construction on a primitive type in the first primitive type set for the first knowledge transfer operation for this time to the final scene construction information based on a received input instruction.

8 . The method according to claim 5 , wherein adjusting the color of each primitive in the scene construction image according to the color matching information corresponding to the acquired image generation condition to obtain the color matching image comprises:

Assignees

• Beijing Wodong Tianjun Information Technology Co Ltd
• Beijing Jingdong Century Trading Co Ltd

Inventors

• Wang Chenyu

Classifications

• G06T11/60

  Creating or editing images; Combining images with text · CPC title

• G06V10/761

  Proximity, similarity or dissimilarity measures · CPC title

• G06V10/56

  relating to colour · CPC title

• Y02D10/00

  Energy efficient computing, e.g. low power processors, power management or thermal management · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title