Beyond Deep Residual Learning for Image Restoration: Persistent Homology-Guided Manifold Simplification

19 Nov 2016  ·  Woong Bae, Jaejun Yoo, Jong Chul Ye ·

The latest deep learning approaches perform better than the state-of-the-art signal processing approaches in various image restoration tasks. However, if an image contains many patterns and structures, the performance of these CNNs is still inferior. To address this issue, here we propose a novel feature space deep residual learning algorithm that outperforms the existing residual learning. The main idea is originated from the observation that the performance of a learning algorithm can be improved if the input and/or label manifolds can be made topologically simpler by an analytic mapping to a feature space. Our extensive numerical studies using denoising experiments and NTIRE single-image super-resolution (SISR) competition demonstrate that the proposed feature space residual learning outperforms the existing state-of-the-art approaches. Moreover, our algorithm was ranked third in NTIRE competition with 5-10 times faster computational time compared to the top ranked teams. The source code is available on page : https://github.com/iorism/CNN.git

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Image Super-Resolution BSD100 - 4x upscaling Manifold Simplification PSNR 27.66 # 22
SSIM 0.7380 # 25
Color Image Denoising CBSD68 sigma50 DnCNN PSNR 28.01 # 5
Image Super-Resolution Set14 - 4x upscaling Manifold Simplification PSNR 28.80 # 30
SSIM 0.7856 # 35
Image Super-Resolution Urban100 - 4x upscaling Manifold Simplification PSNR 26.42 # 27
SSIM 0.7940 # 25

Methods


No methods listed for this paper. Add relevant methods here