Object Oriented Shadow Detection and Removal in Image Having Cast Shadows

A shadow is created when an object lies in the path of a light source. Shadows are cast by the occluding object, or the object itself can be shaded; a phenomenon known as “self-shading”. Due to the difference between the light intensity reaching a shaded region and a directly lit region, shadows are often characterized by strong brightness gradients. While non-shadow regions are illuminated by both direct (e.g., sunlight, flashlight) and diffuse (e.g., skylight, fluorescent, incandescent) light sources, shadow regions are only illuminated by diffuse light. The change between shadow and non-shadow regions is thus not only a brightness difference, but a color one as well. In outdoor scenes, for example, the lit areas are illuminated by sunlight and skylight, while the shadow regions are illuminated by skylight alone, which creates a bluish color cast. This property of shadows, together with the fact that a certain color can exist in both lit and shaded objects, makes them problematic in a number of different computer vision applications such as tracking, object (or people) recognition, and white balancing. In image manipulation or compositing, shadows are often unwanted artifacts that are unavoidable due to image capture conditions (e.g., photographs taken in an urban environment). While shadow compositing and removal has dramatically improved in recent years, automatic shadow detection is still a challenge and requires additional assumptions or information. We have presented a method of shadow detection and removal based on two dimensional chromaticity based intensity ratios which gives good results in detecting and removing of cast shadows in a still image.