Authors: Junghun Cha, Ali Haider, Seoyun Yang, Hoeyeong Jin, Subin Yang, A. F. M. Shahab Uddin, Jaehyoung Kim, Soo Ye Kim, Sung-Ho Bae

Published on: February 08, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.05350

Summary

• What is new: Introduction of DESCAN-18K, a new high-quality large-scale dataset for Descanning, and a novel image restoration model called DescanDiffusion.
• Why this is important: The quality of digitized analog information (documents and images) is often degraded by the printing, storing, and scanning processes, making restoration critical.
• What the research proposes: A new image restoration model, DescanDiffusion, which uses a color encoder and a conditional denoising diffusion probabilistic model to restore scanned images.
• Results: DescanDiffusion significantly outperforms other baselines and commercial products in restoring high-quality content from scanned copies.

Technical Details

Technological frameworks used: DescanDiffusion, a combination of color encoder and conditional denoising diffusion probabilistic model

Models used: Conditional Denoising Diffusion Probabilistic Model (DDPM)

Data used: DESCAN-18K dataset consisting of 18K pairs of original and scanned images with complex degradations

Potential Impact

Digital archiving, online libraries, e-commerce platforms, and companies specializing in document management and digital restoration services could significantly benefit or face disruption.

Want to implement this idea in a business?

We have generated a startup concept here: ClearScan AI.