Shadow detection and removal for moving objects using Daubechies complex wavelet transform

Shadow detection and removal is a challenging problem for several computer vision applications because shadow always makes object misclassified. A number of shadow detection and removal algorithms have been reported, and some of these algorithms require manual calibration in terms of some hypothesis and predefined specific parameters whereas others do not require manual intervention, but fail to give accurate result in various lighting and environmental conditions. This paper introduces a novel method for shadow detection and removal with Daubechies complex wavelet domain. Daubechies complex wavelet transform has been used in the proposed algorithm due to its strong edge detection, approximate shift-invariance as well as approximate rotation invariance properties. For shadow detection, we have proposed a new threshold in the form of coefficient of variation of wavelet coefficients. This threshold is automatically determined and does not require any manual calibration and training. Results of shadow detection and removal from moving objects after applying the proposed method are compared with the those of other state-of-the-art methods in terms of visual performance and number of quantitative performance evaluation parameters. The proposed method is found to perform better than other state-of-the-art methods.