Adaptive algorithms for background estimation to detect moving objects in videos

Anh-Tuan Nghiem 1
Abstract : Detecting foreground pixels is the rst step to detect objects of interest in videos. The objective of this thesis is to propose a new background estimation method to detect foreground pixels. The proposed method can adapt the estimated background to various changes of environment (e.g. changes of illumination or of contextual objects). The proposed background estimation method consists of a new background subtraction algorithm to detect foreground pixels, post-processing algorithms to remove shadow and highlight, and a controller to adapt the background subtraction algorithm to the current scene conditions. The new background subtraction algorithm takes into account the scene characteristics such as dynamic background (e.g. tree leave motion), displacement of contextual objects to improve the foreground detection results. It also proposes a new updating method to better adapt its background representation to the current scene conditions. The algorithms to remove shadow and highlight employ new chromaticity and homogeneity (texture) constraints which are robust to illumination changes. These constraints are constructed based on the illumination model and the camera model. The controller has two adaptation methods for the background subtraction algorithm. The rst method is to selectively update the background representation of the background subtraction algorithm. With this updating method, the background subtraction algorithm can solve various problems such as managing stationary objects, keeping track of objects when they stop moving. The second method is to tune the parameter values of the background subtraction algorithm. To fulfill these two tasks, the controller extensively uses the feedback from the classication task and the information about the background subtraction algorithm and the scene. This method has been validated using the public database ETISEO and one hour video from the project GERHOME.
Document type :
Complete list of metadatas

Cited literature [69 references]  Display  Hide  Download
Contributor : Anh-Tuan Nghiem <>
Submitted on : Monday, July 26, 2010 - 2:39:22 PM
Last modification on : Saturday, January 27, 2018 - 1:31:28 AM
Long-term archiving on : Tuesday, October 23, 2012 - 11:20:22 AM


  • HAL Id : tel-00505881, version 1



Anh-Tuan Nghiem. Adaptive algorithms for background estimation to detect moving objects in videos. Human-Computer Interaction [cs.HC]. Université Nice Sophia Antipolis, 2010. English. ⟨tel-00505881⟩



Record views


Files downloads