A Novel Image Inpainting Method Used for Veneer Defects Based on Region Normalization
Yilin Ge1, Jiahao Chen1, Yunyi Lou1
1College of Mechanical and Electrical Engineering, Northeast Forestry University, Harbin 150040, China.
Sensors (Basel, Switzerland)
|June 24, 2022
Summary
This study introduces a novel deep generative model for defective veneer inpainting, significantly improving wood panel quality. The enhanced method achieves superior image inpainting results, boosting veneer utilization and blockboard grade.
Area of Science:
- Materials Science
- Computer Vision
- Artificial Intelligence
Background:
- Veneer quality is crucial for blockboard production.
- Defective veneers reduce material utilization and product grade.
- Existing inpainting methods may struggle with complex defects.
Purpose of the Study:
- To develop an advanced deep generative model for high-quality veneer inpainting.
- To improve the utilization of defective veneers in blockboard manufacturing.
- To enhance the visual quality and structural integrity of repaired veneer panels.
Main Methods:
- A novel two-phase deep generative network was employed for stable training.
- Region normalization was introduced to address mean/standard deviation inconsistencies and improve convergence.
- A hybrid dilated convolution module was utilized for effective reconstruction of missing veneer areas, mitigating gridding artifacts.
Main Results:
- The proposed method demonstrated superior performance in image inpainting tasks on veneer datasets.
- Achieved a Peak Signal-to-Noise Ratio (PSNR) of 33.11.
- Achieved a Structural Similarity Index Measure (SSIM) of 0.93, outperforming existing methods.
Conclusions:
- The deep generative model effectively improves defective veneer quality and utilization.
- The enhanced inpainting technique offers a viable solution for producing higher-grade blockboards.
- The method shows significant potential for industrial application in wood panel manufacturing.
