Motion foreground detection based on wavelet transformation and color ratio difference

An algorithm is presented for detecting foreground objects with multi-scale wavelet transformation and color ratio difference. Multi-scale wavelet transformation method is used to segment moving objects based on spatial property. Ratio differences between two adjacent pixels in four different directions are used to classify object pixels. RGB color space is selected to segment moving foregrounds and eliminate cast shadows instead of complex color models. The developed approach does not require any complex supervised training phase, manual calibration or hypothesis in removing shadow. Experiments have highlighted that the proposal is efficient to segment foregrounds and suppress shadows.