Ornamental Surface Image Generation With Guided Diffusion Refinement

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for digitizing natural material surfaces for printing and embossing tools require extensive manual preparation and are prone to errors such as hallucinations and logical inconsistencies due to the use of machine learning and artificial intelligence.

Innovation Solution

A computer-implemented method using a cascade of trained machine learning and artificial intelligence models, including diffusion models like Stable Diffusion XL and GAN models, with a guidance input to generate accurate surface images by denoising, refining, and upscaling, reducing errors and improving computing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If machine learning and AI methods are used to generate surface images, then productivity and automation are improved, but reliability deteriorates due to hallucinations and artifacts

Engineering Contradiction:
Improveimage generation speedVSAvoidimage accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements a feedback mechanism where the generated image is processed through a second diffusion model that uses the first image as guidance. This iterative refinement process allows the system to detect and correct hallucinations and artifacts by comparing the generated output against the guidance input, thereby improving reliability while maintaining automation.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent introduces a guidance input as an intermediary element that mediates between the random noise and the final generated image. This guidance input, which may be a sketch, photograph, or existing image, acts as a constraint that prevents the diffusion model from generating unrealistic artifacts while still allowing creative generation, thus resolving the contradiction between automation and reliability.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Manufacturing precision

If extensive manual preparation of material surfaces is performed before digitization, then manufacturing precision is improved, but loss of time increases

Engineering Contradiction:
Improvesurface digitization accuracyVSAvoidpreparation time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent replaces the mechanical preparation process (sawing, sanding, and manual digitization of physical surfaces) with an AI-based diffusion model that can generate surface images directly from textual or visual guidance inputs. This substitution eliminates the need for physical surface preparation while maintaining or improving the quality of the digitized surface representation, thereby resolving the contradiction between precision and time consumption.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

3Reliability

If multiple diffusion models are used in cascade to reduce errors, then reliability is improved, but device complexity increases

Engineering Contradiction:
Improveimage generation accuracyVSAvoidmodel cascade structure
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the image generation process into two distinct diffusion models with specialized functions: the first model generates the initial surface image from noise and guidance, while the second model refines this output by using it as guidance. This segmentation allows each model to be optimized for its specific task, improving overall reliability while keeping the complexity manageable through clear functional separation.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4712030A1Surface design image generation
Publication Date: 2026.03.18 MATTHEWS INTERNATIONAL GMBH
  • EP4712030A1 patent drawingFigure 1
  • EP4712030A1 patent drawingFigure 1A~1D
  • EP4712030A1 patent drawingFigure 2A~2C

AI summary

A computer implemented method for generating an image of an ornamental surface, surface structure and/or a surface model of one or more target materials, the method comprising: • Providing an input to into one or more trained machine learning and/or artificial intelligence models, the input comprising one or more parameters and a guidance input; and • Generating, by the one or more trained machine learning and/or artificial intelligence models, a generated image based at least in part on the one or more parameters and the guidance input, wherein at least one of the one or more trained machine learning and/or artificial intelligence models comprises or is associated with at least one diffusion model.