Static Image Shadow Detection Texture Analysis by Entropy-Based Method

Abstract

For applications such as object identification and object tracking, the primary goal is that the system should be able to track only the object of interest. If it is unable to do, it would lead to false tracking. The images so obtained by the tracking system may have a number of objects with similar shapes and colors. These images can be further outdoor images with illumination conditions so as to cast the shadow of the object. Of these objects present, the objects that map very closely to the object of interest is its shadows. If the image along with its shadow is allowed to enter the system, it may lead to false tracking. This is so because the image acquisition system cannot differentiate between shadow and its object on its own. Therefore, a stage called as shadow detection and elimination stage had to be introduced between the image acquisitions and processing stage in the tracking system. When the shadow detection taxonomy is examined it gives four standard methods of detection based on intensity, color, geometry, and texture. Most of the images map their requirement with first three methods and a very few which does not satisfy the requirements of these three methods adopt the texture-based method of detection. Texture-based method is used mainly in medical fields and other specialized applications. This paper makes an attempt to use texture-based method for shadow detection in case of static images.