Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Adaptive Convolutional Neural Network for Image Super-resolution

Ziang Wu, Jinwei Xie, Xuanyu Zhang, Tao Wang, Yongjun Zhang, Qi Zhu, Chunwei Tian

TL;DR

ADSRNet introduces a dual-path adaptive convolutional design for single-image super-resolution, featuring a heterogeneous upper network to capture contextual and salient information and a symmetrical lower sub-network to reinforce hierarchical relationships. Dynamic and dilated convolutions are embedded within heterogeneous blocks to adapt to varying inputs, while a construction block fuses features via a two-step upsampling and convolutional process. Extensive experiments on DIV2K and four standard SR benchmarks demonstrate state-of-the-art PSNR/SSIM across x2, x3, and x4 scales, with favorable runtime and parameter efficiency. The work advances robust SR across diverse scenes and offers a publicly available implementation for practical deployment in devices.

Abstract

Convolutional neural networks can automatically learn features via deep network architectures and given input samples. However, the robustness of obtained models may face challenges in varying scenes. Bigger differences in network architecture are beneficial to extract more diversified structural information to strengthen the robustness of an obtained super-resolution model. In this paper, we proposed a adaptive convolutional neural network for image super-resolution (ADSRNet). To capture more information, ADSRNet is implemented by a heterogeneous parallel network. The upper network can enhance relation of context information, salient information relation of a kernel mapping and relations of shallow and deep layers to improve performance of image super-resolution. That can strengthen adaptability of an obtained super-resolution model for different scenes. The lower network utilizes a symmetric architecture to enhance relations of different layers to mine more structural information, which is complementary with a upper network for image super-resolution. The relevant experimental results show that the proposed ADSRNet is effective to deal with image resolving. Codes are obtained at https://github.com/hellloxiaotian/ADSRNet.

Adaptive Convolutional Neural Network for Image Super-resolution

TL;DR

ADSRNet introduces a dual-path adaptive convolutional design for single-image super-resolution, featuring a heterogeneous upper network to capture contextual and salient information and a symmetrical lower sub-network to reinforce hierarchical relationships. Dynamic and dilated convolutions are embedded within heterogeneous blocks to adapt to varying inputs, while a construction block fuses features via a two-step upsampling and convolutional process. Extensive experiments on DIV2K and four standard SR benchmarks demonstrate state-of-the-art PSNR/SSIM across x2, x3, and x4 scales, with favorable runtime and parameter efficiency. The work advances robust SR across diverse scenes and offers a publicly available implementation for practical deployment in devices.

Abstract

Convolutional neural networks can automatically learn features via deep network architectures and given input samples. However, the robustness of obtained models may face challenges in varying scenes. Bigger differences in network architecture are beneficial to extract more diversified structural information to strengthen the robustness of an obtained super-resolution model. In this paper, we proposed a adaptive convolutional neural network for image super-resolution (ADSRNet). To capture more information, ADSRNet is implemented by a heterogeneous parallel network. The upper network can enhance relation of context information, salient information relation of a kernel mapping and relations of shallow and deep layers to improve performance of image super-resolution. That can strengthen adaptability of an obtained super-resolution model for different scenes. The lower network utilizes a symmetric architecture to enhance relations of different layers to mine more structural information, which is complementary with a upper network for image super-resolution. The relevant experimental results show that the proposed ADSRNet is effective to deal with image resolving. Codes are obtained at https://github.com/hellloxiaotian/ADSRNet.
Paper Structure (17 sections, 5 equations, 7 figures, 7 tables)

This paper contains 17 sections, 5 equations, 7 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Deep CNNs for Image Super-resolution
  4. Dynamic convolution
  5. Proposed Method
  6. Network Architecture
  7. Loss Function
  8. Heterogeneous Block
  9. Symmetrical Lower Sub-network
  10. Construction Block
  11. Experiments
  12. Training Dataset
  13. Test Datasets
  14. Implementation Details
  15. Network Analysis
  16. ...and 2 more sections

Figures (7)

  • Figure 1: Network architecture of the proposed ADSRNet.
  • Figure 2: Predicted high-quality images from different methods on a same low-resolution image from Set14 for ×2: (a) A HR image, (b)Bicubic, (c) SRCNN, (d) LESRCNN, (e) DCLS, (f) VDSR, (g) DRCN and (h) ADSRNet (Ours).
  • Figure 3: Predicted high-quality images from different methods on a same low-resolution image from B100 for ×3: (a) A HR image, (b)Bicubic, (c) SRCNN, (d) LESRCNN, (e) DCLS, (f) VDSR, (g) DRCN and (h) ADSRNet (Ours).
  • Figure 4: Predicted high-quality images from different methods on a same low-resolution image from B100 for ×3: (a) A HR image, (b)Bicubic, (c) SRCNN, (d) LESRCNN, (e) DCLS, (f) VDSR, (g) DRCN and (h) ADSRNet (Ours).
  • Figure 5: Predicted high-quality images from different methods on a same low-resolution image from U100 for ×4: (a) A HR image, (b)Bicubic, (c) SRCNN, (d) LESRCNN, (e) DCLS, (f) VDSR, (g) DRCN and (h) ADSRNet (Ours).
  • ...and 2 more figures