Vector Pattern Synthesis Using Clustering and Energy Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital rendering systems are limited in synthesizing complex vector patterns, often resulting in poor quality and accuracy due to incorrect structure synthesis between vector elements and overlapping shapes in different depth layers, and require excessive user interaction for simple pattern generation.
Innovation Solution
The system jointly synthesizes and clusters sample distributions to produce robust optimization and reconstruction of output vector patterns using a greedy algorithm for neighborhood matching and an energy-based optimization model to optimize cluster configurations, maintaining structural relationships and shape details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional systems rasterize vector patterns and apply image texture synthesis, then simple vector patterns can be generated, but complex vector patterns cannot be synthesized and quality deteriorates
Solution Approach 1:
The patent segments the vector pattern synthesis process into distinct stages: sampling vector elements from input patterns, clustering them into groups with similar characteristics, and then synthesizing new patterns by recombining these clusters. This segmentation allows the system to handle both simple and complex vector patterns while maintaining structural integrity and design fidelity throughout the synthesis process.
2Productivity
If conventional systems use image texture synthesis to generate vector patterns, then pattern generation is possible, but structural relationships between vector elements are lost
Solution Approach 1:
The patent employs copying mechanisms where vector elements are sampled from input patterns and replicated into clusters. These copied elements maintain their original structural relationships and design characteristics. The synthesis process then recombines these copied elements rather than generating entirely new ones, thereby preserving the original design intent while creating varied output patterns.
3Ease of operation
If conventional systems require user interaction via user interface, then control over pattern generation is possible, but excessive user interaction is required even for simple patterns
Solution Approach 1:
The patent implements self-service functionality where the system automatically performs vector pattern synthesis by sampling, clustering, and recombining vector elements without requiring continuous user intervention. The automated clustering algorithm independently identifies and groups similar vector elements, and the synthesis process autonomously generates new patterns. This self-service capability dramatically reduces user interaction time while maintaining ease of operation through simple interface controls.
Data Source
AI summary
The present disclosure relates to systems, non-transitory computer-readable media, and methods that utilize an optimization model for generating vector patterns with complex vector structures. For example, the disclosed systems iteratively optimize the similarity between local input and output neighborhoods that account for clusters. Specifically, based on an input exemplar vector image, the disclosed systems generate a sample input cluster representation for more robust iterative sample optimization and pattern reconstruction. To illustrate, the disclosed systems optimize output cluster configurations based on input clusters such that the output clusters minimize a shape energy and a link energy (e.g., to better preserve shape and structure details from the original vector pattern in the input exemplar vector image). From the output clusters, the disclosed systems can reconstruct additional vector elements to create a new vector image with a synthetic vector pattern.