RPI ID: 2022-015-201

Innovation Summary:
A deep learning-based image denoising method tailored to specific tasks such as segmentation or classification. Unlike generic denoising, it preserves features critical to downstream analysis. The model adapts its strategy based on the intended application, improving performance in real-world scenarios. It is especially effective in medical and industrial imaging where noise can obscure key details.

Challenges / Opportunities:
Generic denoising often removes important signals. This task-oriented approach ensures preservation of relevant features. It enhances AI performance in noisy environments and can be customized for various imaging modalities. The method reduces diagnostic errors and improves reliability.

Key Benefits / Advantages:
✔ Task-specific optimization
✔ Preserves critical features
✔ Improves AI performance
✔ Adaptable across domains
✔ Reduces false negatives

Applications:
• Medical imaging
• Industrial inspection
• Satellite imaging

Keywords:
image denoising, deep learning, task-oriented, medical imaging, noise reduction, AI optimization

Intellectual Property:
Published US application, 17/954561, US20230099663A1, filed 28-Sep-2022