Phase-based motion estimation in complex environments using the illumination-invariant log-Gabor filter

Yuchao Wang, Weihua Hu, Jun Teng, Yong Xia

Research output: Journal article publicationJournal articleAcademic researchpeer-review

7 Citations (Scopus)


The phase-based motion estimation method is promising because the phase is sensitive to the motion information embedded in the pixel values. Unfortunately, complex environments such as extra textures and illumination changes exist in practice, whereas their effects have not been studied. This study theoretically discovers that the large spatial size and DC component of the filter adversely affect the performance of the motion estimation method under complex environments. The traditional method adopting the Gabor filter with a large bandwidth may suppress extra textures while cause a significant drift in the estimated motion due to the negative frequency and DC components. To tackle this, a log-Gabor-based new motion estimation method is proposed, in which the log-Gabor filter intrinsically without the DC and negative frequency components can extract the local phase and filter out the illumination change. A localization index based on the amplitude map is designed to determine the optimal parameters of the log-Gabor filter and maximize its spatial localization capability. Numerical and experimental examples demonstrate that the proposed method outperforms the traditional Gabor filter in motion estimation under the complex environment.

Original languageEnglish
Article number109847
JournalMechanical Systems and Signal Processing
Publication statusPublished - 1 Mar 2023


  • Complex environment
  • Log-Gabor filter
  • Motion estimation
  • Phase-based method

ASJC Scopus subject areas

  • Control and Systems Engineering
  • Signal Processing
  • Civil and Structural Engineering
  • Aerospace Engineering
  • Mechanical Engineering
  • Computer Science Applications


Dive into the research topics of 'Phase-based motion estimation in complex environments using the illumination-invariant log-Gabor filter'. Together they form a unique fingerprint.

Cite this