[논문리뷰] Noise2Noise - Learning Image Restoration without Clean Data

[논문리뷰] Noise2Noise - Learning Image Restoration without Clean Data

1. Introduction

2. Theoretical Background

z = denoised image

x, y 모두 corrupted image

Averge gradient를 통해 주위의 noisy target으로 부터 clean target을 얻음

3 . Practical Experiments

“clean targets are unnecessary in this application”

3종류 noise의 결과영상