Motion Detection by Using Entropy Image and Adaptive State-Labeling Technique
Meng‐Chou Chang, Yongjie Cheng · 2007
This paper proposes an improved motion detection method based on the entropy image and the adaptive state-labeling algorithm. In our method, a spatio-temporal sliding window is built for each pixel, and the pixels in the sliding window are assigned state labels according to our adaptive state-labeling technique. The state label distribution in the sliding window is used to construct the entropy image, in which a pixel with lower entropy is considered as part of a moving object. In this paper, we have compared our motion detection method with the MRF (Markov random field) based method, the STEI (spatio-temporal entropy image) method, and the DSTEI (difference-based spatio-temporal entropy image) method. Experimental results show that our motion detection method is robust and has lower computational complexity.