AI Paper Atlas

Research map / Image generation / Image restoration

The Devil is in the Upsampling: Architectural Decisions Made Simpler for Denoising with Deep Image Prior

Yilin Liu, Jiang Li, Yunkui Pang, Dong Nie, Pew-Thian Yap

ICCV 2023 · Image generation › Image restoration · approach: Self-supervised image denoising · 18 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use a 2-level hourglass with per-image estimated width instead of the 5-level DIP at high noiseDIP hourglass · Set12 sigma=50 GaussianPSNR: 22.15 -> 24.59Improved
use the estimated under-parameterized network at mild noiseDIP hourglass · CBSD68 sigma=25 GaussianPSNR: 28.93 -> 28.57No clear effect
estimated architecture on real noise versus DIPDIP hourglass · PolyUPSNR: 38.15 -> 38.05No clear effect
remove three upsampling layers and scale width per imageConvDecoder · CBSD68 sigma=25PSNR: 25.19 -> 28.29Improved
same ConvDecoder modification at higher noiseConvDecoder · CBSD68 sigma=50PSNR: 24.36 -> 24.12No clear effect
remove upsampling from the decoder (ablation)MLP-Decoder / ConvDecoder · Image denoising (Set9)denoising PSNRHurt
replace bilinear with transposed convolution upsampling (ablation)decoder · Image denoising (Set9)PSNR over iterationsHurt
use NN upsampler instead of bilinear on a coarse image (last iteration) (ablation)ConvDecoder · Set9 coarse imagePSNR: 31.1 -> 27.8Hurt
use NN upsampler instead of bilinear on a fine image (ablation)ConvDecoder · Set9 fine image (peak)PSNR: 22.9 -> 24.5Improved
use an LPF with about 100dB attenuation in the upsampler (ablation)ConvDecoder · Set9 coarse image (peak)PSNR: 31.3 -> 27.5Hurt
increase network depth (ablation)hourglass DIP · fine-grained imagesSSIMHurt
add skip connections to deep network (ablation)hourglass DIP · fine-grained imagesSSIMImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy