A Novel Motion Detection Method Resistant to Severe Illumination Changes

Authors

Sahar Yousefi, Mohammad T. Manzuri Shalmani, Jeremy Lin, and Marius Staring

Overview

Motion detection can be disrupted by illumination changes, occlusion, camouflage, sudden physical motion, and dynamic textures. This work introduces a motion descriptor designed to remain effective under severe changes in lighting and other difficult video conditions.

Method

The method combines 3D discrete wavelet transform (3D-DWT) features with 3D wavelet leaders to describe motion at each pixel. The resulting feature vectors are grouped using clustering methods such as k-means or Gaussian mixture models (GMM). The paper evaluates the approach for both motion detection and dynamic texture segmentation.

Project links