Background Estimation Based on Temporal Filtering in Background Eigenspace

Yasutomo KAWANISHI  Masayuki MUKUNOKI  Michihiko MINOH  

Publication
D - Abstracts of IEICE TRANSACTIONS on Information and Systems (Japanese Edition)   Vol.J95-D   No.9   pp.1759-1768
Publication Date: 2012/09/01
Online ISSN: 1881-0225
DOI: 
Print ISSN: 1880-4535
Type of Manuscript: PAPER
Category: 
Keyword: 
background estimation,  eigen background,  temporal filtering,  

Full Text(in Japanese): PDF(1MB)
>>Buy this Article


Summary: 
In outdoor scenes, there are usually many background objects moving repeatedly, such as trees or flags, which are swinged by wind. There are also illumination changes caused by sunlight. Therefore it is difficult to estimate the background by preserving the illumination changes and the movements of such background objects. In this paper, we propose a background estimation technique for an outdoor fixed camera using temporal statistics of background variations. In this technique, first, we devide an image into blocks and build a background eigenspace for each block. Then we apply temporal filtering to estimate the current background in the background eigenspace. We experimentally evaluate its effectiveness by comparing the results of the proposed and existing methods for several surveillance cameras.