Shadow detection of static images by innovative thresholding approach

Computer vision systems majorly work on the basis of capturing three dimensional scenes. The scenes however have two dimensional images or videos that are sent for processing. The said images and the videos always have an object of interest. The scenes which are captured have a number of objects along with the object of interest. of all these objects, the ones which are not required are ignored or considered as the background. Consider applications like tracking and object identification, where the object of interest has to be identified. The computer vision system which captures these scene, has along with other objects, Object of interest and also their shadows. If only the object of interest is considered then its shadow having a shape almost similar to that of the object and further if the object is dark then the matching is at shape as well as colour level. There are always chances that the tracking and the object identification system which works on colour and shape, may detect the shadow in place of the object. This is called as false tracking or false detection respectively. The limitations of the computer vision system is that it cannot differentiate between the objects and its shadows, so a preprocessing stage has to be introduced between the computer vision system and processing stage called as the shadow detection and elimination system. The algorithm for the same maps in four working domain namely intensity, colour, texture and geometry. This paper highlights the colour based method based on global thresholding used for shadow detection with suggested modifications at the local level. This is the work which is presented in this paper. It further uses the qualitative and quantitative evaluation methods to compare the modified method with existing global thresholding method.