A moving object extraction algorithm based on the modified codebook model
Tingshu Lu, Zhongxin Qu, Xiaojiong Liu, Mingjiang Wang · 2015
The extraction of moving targets in complex background scenes is an important part of computer vision applications. The basic method to extract the interest object is to build time series models of pixels. However, these models are often complex and also consume memories. In order to reduce the computational complexity and to streamline computing time, an improved model base on the conventional codebook model is proposed in this paper. Our improved background model is constructed with codebooks consists of codewords. The codewords are encoded based on a box model corresponding to the change scopes of pixels. At the pixel modeling stage, by using blocks surrounding the pixel area to initialize the codebook, and adjusting the position of access codewords, the modeling time is reduced. Our experiments show that the improved codebook algorithm proposed has the higher robustness and accuracy than the conventional codebook algorithm.