Applications

  • Follow this link to see some case studies:
    1. A typical videosurveillance
    2. Small objects detection: ViBe is very precise for the contour of objects because there is no spatial filtering. This is particular useful for an efficient detection of small objects. 
    3. Simultaneous use of ViBe on depth map and on RGB for a better delineation of objects
    4. ViBe’s performances in noisy images
    5. Post-processing tool to segment objects.
    6. Inpainting with ViBe
    7. ViBe on range (depth) maps


  • Evaluation of background subtraction techniques for video surveillance. "Considering these aspects, ViBE is a strong favorite, since it is simple and almost parameterless." click here S. Brutzer, B. Hoferlin, and G. Heidemann, In IEEE International Conference on Computer Vision and Pattern Recognition (CVPR), pages 1937-1944

  • Image-based people detection. "During the past 10 years, there have been many breakthroughs in people detection research. First, …. Second, Primesense and Microsoft released the Kinect depth camera in 2010, …Third, the accuracy and computational cost of many decades old background subtraction has been improved by (Barnich & Van Droogenbroeck, 2011) with ViBE in 2011. " T. Tikkanen., 2013

  • A conservative scene model update policy.  “The ViBe model is unique in that it is the first and only scene model that uses a completely stochastic maintenance algorithm to integrate new information into the system. We implemented ViBe and immediately observed its superiority to several other well known scene modeling techniques.”. This work was supported in part by the U.S. Army Research Laboratory and the U.S. Army Research Office under grant W911NF- 08-1-0293. N. Mould and J. Havlicek, Santa Fee, New Mexico, USA, April 2012.

  • Real-time Aerial Targets Detection Algorithm Based Background Subtraction. "We need a system which is good at anti-jamming, fast response and stability tracking capability. To accomplish this goal, we must detect, segment and track aerial objects steady and automatically. (…) Based on this, we proposed to base the algorithm on ViBE. (...) The proposed algorithm in this paper is simple and effective, has high detection accuracy and strong tracking stability, and also the higher real-time performance and robustness. So, the use of background subtraction for aerial targets detection and tracking has a certain practicality. " Zheng, Wu, Bakhdavlatov, Qu, Li, Yuan - College of Electronic and control Engineering, Beijing University of Technology.
 
  • Quantitative Performance Analysis of Object Detection Algorithms on Underwater Video Footage.  “The ViBe algorithm excelled in nearly all the videos, both in terms of DR (Detection Rate) and FAR (False Alarm Rate)". I. Kavasidis and S. Palazzo, Italy, Nov. 2012.   

  • See www.changedetection.net for a comparative benchmarking of ViBe (and ViBe+) on 31 sequences. Please note that ViBe+ has the best precision of all the techniques. SGMM-SOD that ranks #1, is at least 40 times slower than ViBE and applies a series of pre- and post-processing filters while ViBE doesn’t. PBAS, that ranks #2, is actually based on ViBE but has adapted its parameters to maximize the global score across all sequences.

  • Real-time Implementation of the ViBe Foreground Object Segmentation Algorithm. "This paper describes the implementation of the ViBe background generation algorithm in FPGA. (…). The results show that, using an appropriate hardware platform, with a fast external RAM, allows implementing in a pipeline manner a quite complex video stream analysis algorithm in real-time. The proposed system enables processing of a colour video stream with a resolution of 640 × 480 and 60 frames per second. The module can be used in advanced, automated video surveillance systems and other application with require a reliable foreground mask and real-time image processing." Kryjak, Gorgon - AGH University of Science and Technology, Krakow, Poland

  • A fast on-line boosting tracking algorithm based on cascade filter of multi-features. “…Considering morphologic change of moving object and enhancing the robustness of window selecting, we propose Object Similarity Statistic (OSS) model and use it to filter windows. This idea comes from the ViBe background modeling method.” HU Song, SUN Shui-Fa*1, MA Xian-Bing, QIN Yin-Shi, LEI Bang-Jun - 3rd International Conference on Multimedia Technology


  • Background subtraction using spatiotemporal condition information.ViBe is the state of the art background subtraction method, which has achieved good performance both on accuracy and computational cost.” Bin Wang∗, Yu Liu, Wei Xu, Wei Wang, Maojun Zhang, Accepted 26 August 2013


  • Spatial Mixture of Gaussians for dynamic background modeling“The total number of errors from all sequences (TE) for each algorithm is also shown. … ViBe is one of the best background sub- traction algorithms currently in literature.” Sriram Varadarajan, Paul Miller and Huiyu Zhou - 2013 10th IEEE International Conference on Advanced Video and Signal Based Surveillance.


  • Statistical algorithm for traffic flow based on ViBe. "A quick statistical algorithm for traffic flow was proposed. The improved ViBe algorithm was used to model the background, Then the pixels of video for background was judged, and the background pixels was only updated in order to extract the motion objects from the video. An appropriate human visual habit of YUV color space was adopted to eliminate shadows on the interference of traffic statistics. A pixel level and frame level based on the combination of dynamic adaptive background updated algorithm was presented to meet illumination change. And the improved second round scanning algorithm was used for vehicle division. Then the test window on the road opened up according to the ViBe model and its updated frequency was used to judge whether a vehicle into the observation window, thereby the traffic statistics was realized. Compared with the traditional algorithm, the real—time and robustness of this system is higher." Jiang Jianguo‘’2 Wang Ta01 Qi Meibinl’2 An Hongxinl -2012


  • ViBeExt: the extension of the universal background subtraction algorithm for distributed smart camera“A fast and efficient improvement of ViBe algorithm based on the edge characteristic info and neighborhood mean filter is proposed in order to improve (...) ViBe. This method, ViBeExt, improved the accuracy and robustness of the high speed background subtraction method, ViBe, at a very limited additional cost. The experimental results show that the proposed algorithm can get accurate moving objects in complex scenes. Thus ViBeExt is especially suited for distributed smart camera.”...“Specific strengths of ViBeExt include fast ghost suppression and intrinsic resilience to sudden changes in light. ViBeExt is the only technique that manages to combine a low rate of the number of false positives with both a precise and accurate detection of the foreground pixels.”Fang Zhu, Jiangsu Security and Video Surveillance Engineering Research Center Nanjing, Ping Jiang, Zhaobin Wang - Science and Technology Department of Police, Nanjing


  • A Shadow Removal Algorithm for ViBe in HSV Color Space. "Recently, more and more attention was paid on the ViBe foreground extraction algorithm for its simplicity and high speed. (…). In this paper, a new shadow removal algorithm for ViBe in HSV (Hue, Saturation, Value) color space is proposed." QIN Yin-Shi, SUN Shui-Fa*1, MA Xian-Bing, Hu Song, LEI Bang-Jun


  • Moving object detection based on improved VIBE and graph cut optimization. "In this paper, (…) we present a novel moving object detection method based on improved VIBE and graph cut method from monocular video sequences." J. Dou, J. Li, Optik - Int.J. Light Electron Opt. (2013), http://dx.doi.org/10.1016/j.ijleo.2013.04.106


  • Combining Patch Matching and Detection for Robust Pedestrian Tracking in Monocular Calibrated Cameras. "There are innumerous background removal algorithms proposed in the literature and we have chosen ViBE due to its good performance for surveillance videos and its capability of online adapting the background model." Führ, G., Jung, C.R., Combining Patch Matching and Detection for Robust Pedestrian Tracking in Monocular Calibrated Cameras, Pattern Recognition Letters (2013), doi: http://dx.doi.org/10.1016/ j.patrec.2013.08.031


  • Fast Pedestrian Detection Based on Sliding Window Filtering. “We choose the ViBe model to perform background subtraction. Compared to other background models, this model is appropriate for our method for four reasons: (1) It can fast separate foreground pixels from the background by combining random process to the background subtraction process. (2) It only needs a single frame to initialize the model which is necessary to start pedestrian detection without latency. (3) It uses information of neighborhood pixels to update the model which will achieve good accuracy to get foreground pixels. (4) The objects will be treated as foreground if they stop moving for a short period of time. (It is hard and not common for pedestrians to keep absolutely static for a long time especially in video surveillance, so no additional steps are needed to handle this case.). Feidie Liang, Dong Wang, Yang Liu, Youcheng Jiang, and Sheng Tang.


  • Ma-Th Algorithm for People Count in a Dense Crowd and their Behaviour Classification“In this paper, ViBe the universal background subtraction method is used to extract the foreground from video sequences, which outperforms the existing GMM.” - Department of Electronics & Communication Engineering Thiagarajar College of Engineering Madurai


  • Enhancing the MST-CSS Representation Using Robust Geometric Features, for Efficient Content Based Video Retrieval (CBVR). “Initially, the moving video object is segmented using ViBE” - Chiranjoy Chattopadhyay and Sukhendu Das, Visualization and Perception Lab Department of Computer Science and Engineering, Indian Institute of Technology, Madras, India


  • An Evaluation of Pixel-Based Methods for the Detection of Floating Objects on the Sea Surface“In this paper we have applied a selection of background subtraction algorithms to the problem, and we show that the recent algorithms such as ViBe and behavior subtraction, which take into account spatial and temporal correlations within the dynamic scene, significantly outperform the more conventional parametric techniques, with only little prior assumptions about the physical properties of the scene.” Alexander Borghgraef, Olivier Barnich, Fabian Lapierre, Marc Van Droogenbroeck, Wilfried Philips, and Marc Acheroy - EURASIP Journal on Advances in Signal Processing Volume 2010, Article ID 978451, 11 pages doi:10.1155/2010/978451


  • Event Detection in Underwater Domain by Exploiting Fish Trajectory Clustering. “In this work we used the approach proposed by ViBE” – Palazzo, Spampinato Department of Electrical, Electronical and Computer Engineering University of Catania Catania, Italy; Cigdem Beyan IPAB, School of Informatics University of Edinburgh Edinburgh, UK


  • Quantitative Performance Analysis of Object Detection Algorithms on Underwater Video FootageViBe excelled in nearly all the videos, both in terms of DR and FAR.” – Kavasidis, Palazzo - Department of Electrical, Electronical and Computer Engineering University of Catania Catania, Italy


  • An improved algorithm for fast detection of target. A method for fast target detection based on video sequence is proposed. ViBe is used to model background pixel. (…). Image morphology and YUV color space were adopted to eliminate the interference of the shadows on target detection. On this basis, the morphological image processing technique was used to improve the proposed algorithm, and the practicability of the algorithm was verified through a number of comparative experiments.Wu Zhenrong, Mao Zheng, Qu Jingsong, Li Hongyan, Yuan Jianjian (College of Electronic and control Engineering, Beijing University of Technology, Beijing 100124, China)