Scanner-Guided Image Processing for Precise Generative AI Composition

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing AI-Generated Content systems struggle to properly provide instructions for image generation regarding object direction and background composition using character strings as prompts.

Innovation Solution

An image processing apparatus that reads an original, generates image data using generative AI, and receives the generated data, allowing for the inclusion of composition and background information in the instruction, enabling proper image generation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a prompt formed by a character string of natural language is used as an instruction for image generation, then the user can input keywords to generate images, but it is impossible to properly provide instructions concerning composition including direction and background

Engineering Contradiction:
Improveease of promptingVSAvoidinstruction precision
Core Design Contradiction:
Ease of operationVSManufacturing precision

Solution Approach 1:

The patent introduces an image processing apparatus as an intermediary between the user's simple text prompt and the generative AI system. This apparatus analyzes the prompt, supplements it with appropriate composition and background information, and transforms it into a detailed instruction set that the generative AI can properly execute. The intermediary resolves the contradiction by bridging the gap between ease of prompting and instruction precision.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The image processing apparatus performs preliminary analysis and preparation of the prompt before it is sent to the generative AI system. It pre-processes the natural language prompt by identifying key elements, determining appropriate compositions, and selecting suitable backgrounds in advance. This preliminary action ensures that when the generative AI receives the instruction, it already contains all necessary details for precise image generation.

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If simple text prompts are used for image generation, then the system is easy to use, but the generative AI cannot properly understand composition and background requirements

Engineering Contradiction:
Improveprompt flexibilityVSAvoidimage generation accuracy
Core Design Contradiction:
Adaptability or versatilityVSReliability

Solution Approach 1:

The image processing apparatus serves as a mediator that maintains the flexibility of text-based prompting while ensuring reliable image generation. It accepts various forms of user input (simple keywords, natural language prompts) and transforms them into structured, reliable instructions that guarantee accurate image generation according to the user's intentions.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The system changes the parameters of the prompt by transforming simple text inputs into enriched instruction sets with multiple parameters (composition, background, direction, lighting, etc.). This parameter transformation maintains the flexibility of natural language input while ensuring the generative AI receives comprehensive instructions for accurate image generation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS20250238970A1Image processing apparatus capable of properly providing instruction for image generation to generative artificial intelligence, method of controlling image processing apparatus, and storage medium
Publication Date: 2025.07.24 CANON KK
  • US20250238970A1 patent drawing
  • US20250238970A1 patent drawing
  • US20250238970A1 patent drawing

AI summary

An information processing apparatus capable of properly providing an instruction for image generation to a generative artificial intelligence (AI). A scanner unit reads an original. A controller causes the generative AI to generate, based on scan image data acquired by reading the original by the scanner unit, AI image data. The controller receives the second image data generated by the generative AI.