Virtual Defect Image Blending to Reduce Edge Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating virtual defect images result in visible edges when defects are cropped and synthesized into normal images, degrading the learning performance of artificial intelligence models.
Innovation Solution
A system and method that segment defects on a pixel basis, apply transformation methods like rotation, scaling, and image blending to create a virtual defect image, adjusting pixel values to match the normal image edges, thereby reducing edge artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a defect portion of the defect image is cropped and synthesized into a normal image, then a virtual defect image is generated, but an edge of the defect becomes visible and learning performance of AI models degrades
Solution Approach 1:
A mask image is introduced as an intermediary element to control the synthesis process. The mask image defines the defect region and enables precise control over which pixels are synthesized, preventing visible edges by ensuring smooth transitions at boundaries. The mask acts as a mediator between the defect image and normal image during the synthesis operation.
Solution Approach 2:
Different regions of the synthesized image are treated differently: the defect region uses pixels from the defect image while the surrounding normal region uses pixels from the normal image. This local differentiation ensures that edge portions have appropriate pixel characteristics matching the surrounding area, reducing visible artifacts while maintaining defect visibility in the center.
2Adaptability or versatility
If defects are segmented and synthesized into normal images, then virtual defect images are created for AI training, but the edge portions show artificial boundaries that reduce image naturalness
Solution Approach 1:
The mask image serves as an intermediary that enables flexible defect synthesis while maintaining image naturalness. By using the mask to guide the synthesis process, the system can adaptively select pixels from either the defect or normal image based on location, creating versatile virtual defects that blend naturally into the background.
Solution Approach 2:
The synthesis process dynamically changes pixel selection parameters based on spatial position relative to the mask. Pixels within the masked defect region are selected from the defect image, while pixels outside are selected from the normal image. This parameter change approach ensures smooth transitions and maintains image naturalness while providing synthesis flexibility.
Data Source
Figure 1~3
Figure 4~5
Figure 6~7
AI summary
The present disclosure relates to a system and method for generating a virtual defect image. According to the present disclosure, a system for generating a virtual defect image may include an input unit configured to receive a normal image and a defect image comprising at least one defect, an extraction unit configured to segment and extract a defect, represented by a plurality of pixels. in the defect image, a preprocessing unit configured to apply a preset transformation method on the defect so as to transform the defect into a transformed defect, and a synthesis unit configured to synthesize the transformed defect into the normal image and adjust the transformed defect.