Conference Paper

Moving cast shadow detection from a Gaussian mixture shadow model

Dept. of Electr. & Comput. Eng., Laval Univ., Que., Canada
DOI: 10.1109/CVPR.2005.233 Conference: Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on, Volume: 2
Source: DBLP

ABSTRACT Moving cast shadows are a major concern for foreground detection algorithms. Processing of foreground images in surveillance applications typically requires that such shadows have been identified and removed from the detected foreground. This paper presents a novel pixel-based statistical approach to model moving cast shadows of non-uniform and varying intensity. This approach uses the Gaussian mixture model (GMM) learning ability to build statistical models describing moving cast shadows on surfaces. This statistical modeling can deal with scenes with complex and time-varying illumination, and prevent false detection in regions where shadows cannot be detected. Gaussian mixture shadow models (GMSM) are automatically constructed and updated over time, are easily added to GMM architecture for foreground detection, and require only a small number of parameters. Results obtained with different scene types show the robustness of the approach.

  • Journal of Signal and Information Processing 01/2011; 02(02):72-78. DOI:10.4236/jsip.2011.22010
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Many computer vision applications dealing with video require detecting and tracking of objects. When the objects of interest have a well defined shape, template matching or more sophisticated classifiers can be used to directly segment the objects from the image. These techniques work well for well defined objects such as vehicles but are difficult to implement for no rigid objects such as human bodies. Shadows cause serious problems while segmenting and extracting objects, due to the misclassification of shadow points as foreground / object. Shadows can cause object merging, object shape distortion and even object losses (due to the shadow cast over another object(s)). Although the rapid development of computer vision essentially requiring shadow detection and extraction methodologies, still this domain is in infant stage. Diverse information that characterizes shadows is exploited and in many cases such information is combined or used in a different way. This makes very difficult to classify in a unique manner the shadow detection methods. This research provides a comprehensive survey of recent research methods and techniques classified by many researchers. Through many classification categories are present, no report completely evaluate the methodologies. This research presents the survey of various algorithms and their applicability. The evaluation in both quantitative and qualitative methods shows the performance comparison each methodology. Certainly this paper will help the prospective researchers in shadow domain.
  • [Show abstract] [Hide abstract]
    ABSTRACT: We present a two-stage method to accurately segment single or multiple moving objects and their shadows, especially when the moving objects have similar chromaticity and intensity to their shadows or when they are immersed in the shadows of other moving objects. Our algorithm first detects potential shadows via brightness ratios at each motion region, which is already separated from the background of an image sequence. Movement patterns are then applied to optimize the regions of moving objects and their shadows. We conducted experiments using our captured image sequences and public videos of Highway I and II to verify our method. The results demonstrate the method's efficiency quantitatively and qualitatively in comparison with ground truth and several advanced methods.
    2014 IEEE Intelligent Vehicles Symposium (IV); 06/2014


Available from