Multi-Scale Correlation for Deep Homography Estimation

Nan KE, Zhaowei SHANG, Lingzhi ZHAO, Yingxin WANG, Mingliang ZHOU*

*Corresponding author for this work

Research output: Journal PublicationsJournal Article (refereed)peer-review

Abstract

In this paper, we propose a novel multi-scale correlation network (MSCNet) for homography estimation from coarse to fine. First, we extract multi-scale features through a siamese network to generate global and local correlations from feature maps of different scales. Second, we use a group dilated deconvolution block to capture global mapping by increasing the receptive fields in terms of different levels. Third, we employ the channel and spatial attention mechanism to achieve local refinement for small displacements. Finally, we adopt a knowledge distillation strategy to lightweight our model while maintaining relatively high estimation performance. Experimental results on Microsoft Common Objects in Context (MSCOCO) dataset show that our proposed MSCNet outperforms the state-of-the-art approaches in terms of accuracy and parameter count.

Original languageEnglish
Article number2250145
JournalJournal of Circuits, Systems and Computers
Volume31
Issue number8
Early online date31 Jan 2022
DOIs
Publication statusPublished - 30 May 2022
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2022 World Scientific Publishing Company.

Keywords

  • dilated deconvolution
  • Homography estimation
  • knowledge distillation
  • multi-scale correlation

Fingerprint

Dive into the research topics of 'Multi-Scale Correlation for Deep Homography Estimation'. Together they form a unique fingerprint.

Cite this