AI Paper Atlas

Research map / Image generation / Image restoration

A Unified Approach of Multi-Scale Deep and Hand-Crafted Features for Defocus Estimation

Jinsun Park, Yu-Wing Tai, Donghyeon Cho, In So Kweon

CVPR 2017 · Image generation › Image restoration · approach: Inverse imaging with learned priors · 127 citations (OpenAlex)

Check 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
concatenate deep and hand-crafted features instead of using the deep feature alone (ablation)Defocus feature classifier · 11-class defocus feature classificationaccuracy (%): 89.38 -> 94.16Improved
use multi-scale patch extraction instead of single-scale extraction (ablation)f_B defocus feature classifier · 11-class defocus feature classificationaccuracy (%): 91.00% -> 94.16Improved
add the gradient feature to [f_D, f_S, f_C] (ablation)Defocus feature classifier · 11-class defocus feature classificationaccuracy (%): 93.25% -> 94.16Improved
add the DCT feature to [f_G, f_S, f_C] (ablation)Defocus feature classifier · 11-class defocus feature classificationaccuracy (%): 91.10% -> 94.16Improved
combine deep and hand-crafted features rather than use either feature family alone (ablation)Defocus map estimation · Blur detection datasetdefocus map qualityImproved
propagate sparse estimates using an edge-preserving smoothed image rather than the color-line-model priorDefocus map estimation · Blur detection datasetsegmentation qualityImproved
evaluate the proposed method instead of Shi et al.'s methodDefocus estimation · Shi et al. datasetRelOrder: 0.1393 -> 0.1572Improved
add random classification seed points in large homogeneous regions (ablation)Sparse defocus map propagation · Large homogeneous image regionsdefocus accuracyImproved

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