Generative Graphic Editing for Readable Contrast-Preserving Designs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing digital design systems struggle to accurately preserve readability and visibility of design assets within digital designs while maintaining a cohesive aesthetic, as they either focus solely on color recommendation or layout recommendation, leading to designs that are either difficult to read or blend too well with the background.

Innovation Solution

The design harmonization system uses a diffusion neural network to generate and modify digital designs by relocating and recoloring design assets, incorporating contrast data and adaptive strength to ensure visibility and readability, employing techniques like prompt cleaning, region proposal, and content-aware correction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Stability of the object's composition

If conventional recoloring models (GANs/BERTs) are used to match color palettes, then color coherence is improved, but readability and visibility of design assets deteriorate

Engineering Contradiction:

Improvecolor coherenceVSAvoidreadability

Core Design Contradiction:

Stability of the object's compositionVSMeasurement precision

Solution Approach 1:

The system applies different processing strategies to different regions: the background image undergoes color palette extraction and matching, while design assets undergo selective recoloring based on their relationship to the background. This local differentiation allows color coherence in the background while preserving readability of foreground elements through contrast maintenance.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system dynamically adjusts colors of design assets by extracting color palettes from the background and applying transformation rules that modify asset colors to achieve both harmony with the background and sufficient contrast for readability. This involves chromatic adaptation while preserving luminance relationships.

Inventive Principle:

Principle #32Color changes

2Stability of the object's composition

If design assets are recolored to match background palette, then aesthetic cohesion is improved, but visibility and distinguishability worsen

Engineering Contradiction:

Improveaesthetic cohesionVSAvoidvisibility

Core Design Contradiction:

Stability of the object's compositionVSReliability

Solution Approach 1:

The system modifies color parameters of design assets selectively, adjusting hue and saturation to achieve aesthetic cohesion while preserving or enhancing luminance contrast. By changing specific color parameters rather than applying uniform recoloring, the system maintains visibility while achieving cohesion.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The system evaluates the relationship between design assets and background, using feedback from color palette analysis to determine appropriate recoloring intensity and direction. This feedback mechanism ensures that recoloring achieves aesthetic cohesion without compromising visibility by continuously assessing contrast relationships.

Inventive Principle:

Principle #23Feedback

3Measurement precision

If generative models are used to relocate and recolor design assets, then readability is improved, but system complexity increases

Engineering Contradiction:

ImprovereadabilityVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system employs a multi-functional framework where a single generative model architecture performs multiple tasks: extracting color palettes from backgrounds, determining optimal asset placements, and generating recoloring instructions. This universal approach improves readability while managing complexity by consolidating functions rather than requiring separate specialized systems.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system introduces an intermediary color palette representation that mediates between the background image and design assets. This intermediary structure simplifies the complex interaction by providing a standardized color framework that guides both relocation and recoloring operations, reducing overall system complexity while improving readability outcomes.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12518450B2Generating graphic designs by exploiting contrast through generative editing

Publication Date: 2026.01.06 ADOBE INC

AI summary

The present disclosure relates to systems, methods, and non-transitory computer readable media for generating digital designs utilizing a diffusion neural network to preserve readability and design composition while modifying image content background images and design assets. In some embodiments, the disclosed systems access a text prompt defining visual attributes of a digital design. Furthermore, the disclosed systems generate a modified text prompt by replacing chromatic information within the text prompt. Additionally, the disclosed systems determine an adaptive strength for a diffusion neural network from the text prompt. Also, the disclosed systems generate a modified digital design utilizing the diffusion neural network to process the modified text prompt according to the adaptive strength.