Raster Image Vectorization Using Graph-Based Object Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting raster images to vector images are inefficient, requiring manual tracing due to poor quality and accuracy in edge detection, and are sensitive to noise, leading to blurry or distorted results.
Innovation Solution
A computer-implemented method using preprocessing, edge detection, and graph schema generation to enhance and ungroup raster images, leveraging feature learning models and nested profiling knowledge graphs for accurate object and edge identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manual tracing is used to convert raster images to vector images, then the quality of the resulting vector images can be improved, but the time consumption and cost increase significantly
Solution Approach 1:
The system performs preliminary actions by preprocessing the raster image (denoising, smoothing, grayscale conversion, border addition) before the main vector conversion process. This prepares the image data in advance to facilitate more accurate and efficient edge detection and object identification, reducing the need for manual intervention while maintaining high quality results.
Solution Approach 2:
The patent introduces an intermediary computational process between the raster image and the final vector image. This intermediary system uses multiple processing stages (preprocessing, edge detection, graph schema generation, ungrouping) to bridge the gap between raster and vector formats, achieving automated conversion with quality comparable to manual tracing but without the time cost.
2Loss of time
If automated edge detection is used to convert raster images to vector images, then the time consumption is reduced, but the quality and accuracy of the resulting vector images deteriorate due to noise sensitivity and blurry results
Solution Approach 1:
The conversion process is segmented into multiple distinct stages: preprocessing (denoising, smoothing), grayscale conversion, border addition, edge detection, graph schema generation, and ungrouping. Each stage addresses specific aspects of the conversion problem, with preprocessing stages specifically designed to handle noise and improve edge detection accuracy, thereby maintaining high quality while achieving automated processing.
Solution Approach 2:
Preliminary preprocessing actions (denoising, smoothing, grayscale conversion, border addition) are performed before edge detection to prepare the image data. These preliminary actions remove noise and enhance edges, ensuring that the subsequent automated edge detection produces high-quality results without manual intervention.
3Productivity
If simple conversion methods are used, then the processing speed is improved, but the ability to accurately identify and separate objects in the image deteriorates
Solution Approach 1:
The patent replaces manual mechanical tracing with an automated computational system. This substitution uses computer-based image processing algorithms (preprocessing, edge detection, graph schema generation) to perform object identification and separation, achieving both high processing speed and accurate object identification without manual intervention.
Solution Approach 2:
An intermediary graph schema is generated between the processed image and the final vector output. This graph schema represents objects and their relationships, enabling the system to accurately identify and separate objects while maintaining processing efficiency. The graph schema acts as a mediator that structures the image data for accurate vector conversion.
Data Source
AI summary
Method, system, and computer-readable storage media for image conversion. A first image in a raster format is preprocessed to form a second image. Further, the second image is converted into grayscale to create a third image. Furthermore, the third image is denoised and smoothened to create a fourth image. A fifth image is created based on the fourth image by at least adding borders to objects in the third image. Further, a layering of the objects is first determined within the fifth image. Also, the objects in the fifth image are highlighted to create a sixth image. A graph schema is generated based on the fifth and/or sixth image. Based on the graph schema, edges of the objects within the sixth image are identified. Further, the sixth image is ungrouped based on the identified edges of the objects, the ungrouping generates an individual vector image for each of the objects.