Multi-Feature Background Modeling Algorithm Based on Improved Census Transform

Fig. 1. Pixel value in 3×3 window. (a) Fields in the window with the same pixels and the different center pixels; (b) center pixel is the original gray value; (c) center pixel is the average gray value of each pixel value of the window; (d) center pixel is the median value of each pixel value in the window

Download full size

Fig. 2. Neighborhood pixel value change. (a) Video sequence; (b) rectangular area pixel value change; (c) center pixel is 153, the t-th frame Census transform; (d) center pixel is 153, the (t+1) th frame Census transform; (e) center pixel is 153, the (t+2)th frame Census transform

Download full size

Fig. 3. Census template. (a) Census template pixel value; (b) t1 frame pixel value; (c) t2 frame pixel value; (d) Census template transform value; (e) I(p')-I(p't1); (f) I(p')-I(p't2)

Download full size

Fig. 4. Different dynamic areas in the video. (a) Continuous 100 frames; (b) complexity of areas A and B

Download full size

Fig. 5. Light mutation scenes

Download full size

Fig. 6. Processing results comparison of five algorithms

Download full size

Table 1. Light mutation scene accuracy verification

View table

Table 1. Light mutation scene accuracy verification

Method	755th frames			801st frames			827th frames
Method	R	P	P_PWC	R	P	P_PWC	R	P	P_PWC
CodeBook	0	0	0	0.89	0.34	51	0.78	0.5	55
MOG	0	0	0.35	0.82	0.38	41	0.71	0.47	59
PBAS	0	0	0.042	0.97	0.33	55	0.92	0.53	48
ViBe	0	0	0.22	0.79	0.33	48	0.68	0.47	59
Proposed	0	0	0	0.25	0.9	21	0.32	0.95	38

Table 2. Processing speed comparison of five algorithms

View table

Table 2. Processing speed comparison of five algorithms

Method	WavingTrees		Bootstrap		Highway		Fountain02
Method	M_MT	H_HT	M_MT	H_HT	M_MT	H_HT	M_MT	H_HT
CodeBook	0.59	0.05	0.62	0.05	1.25	0.08	1.39	0.10
MOG	2.31	0.29	2.30	0.26	8.50	1.30	8.98	1.63
ViBe	0.09	0.03	0.11	0.03	0.45	0.14	0.49	0.18
PBAS	0.08	0.37	0.09	0.36	0.31	0.58	0.46	0.61
Proposed	0.07	0.08	0.06	0.08	0.36	0.28	0.35	0.39

Table 3. Common scene comparison experiment

View table

Table 3. Common scene comparison experiment

Method	Bootstrap			TimeOfDay			WavingTrees			Highway
Method	R	P	P_PWC	R	P	P_PWC	R	P	P_PWC	R	P	P_PWC
CodeBook	0.374	0.626	11.3	0.366	0.998	4.32	0.766	0.727	16.00	0.703	0.891	4.81
MOG	0.290	0.626	11.7	0.494	0.310	11.00	0.695	0.693	18.70	0.748	0.962	3.55
PBAS	0.386	0.568	12.0	0.404	0.981	4.11	0.997	0.677	14.60	0.807	0.98	2.63
ViBe	0.397	0.525	12.8	0.315	0.998	4.67	0.654	0.688	19.70	0.706	0.937	4.29
Proposed	0.588	0.576	11.2	0.435	0.981	3.90	0.962	0.878	5.25	0.845	0.972	2.25

Table 4. Complex scene comparison experiment

View table

Table 4. Complex scene comparison experiment

Method	Canoe			Fountain02
Method	R	P	P_PWC	R	P	P_PWC
CodeBook	0.530	0.341	15.90	0.926	0.577	1.570
MOG	0.295	0.495	10.70	0.609	0.707	1.340
PBAS	0.949	0.410	15.10	0.582	0.589	1.720
ViBe	0.711	0.663	6.91	0.608	0.684	1.400
Proposed	0.833	0.729	5.070	0.809	0.755	0.944

Tools

Get Citation

Copy Citation Text

Zhicheng Guo, Jianwu Dang, Yangping Wang, Jing Jin. Multi-Feature Background Modeling Algorithm Based on Improved Census Transform[J]. Acta Optica Sinica, 2019, 39(8): 0815003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Machine Vision

Received: Mar. 6, 2019

Accepted: Apr. 8, 2019

Published Online: Aug. 7, 2019

The Author Email: Jianwu Dang (lzjdgr@163.com)

DOI:10.3788/AOS201939.0815003

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology