Commentary Paper on "Shadow Removal in Indoor Scenes"
Michal Sofka · 2008
The technique proposed in this paper combines three algorithms for shadow removal in indoor scenes. First algorithm takes an advantage of the assumption on the chromaticity consistency: the ambient light chromaticity is approximately the same as the chromaticity of the diffuse light. Magnitude of differences between chromaticity values of shadow and non-shadow regions for different hue values are different for RGB and HSV color spaces so both spaces are used in conjunction by setting thresholds on these differences. Second algorithm takes a potential shadow region and applies a threshold on the intensity reduction of each pixel and its neighboring pixels. It is ensured that the shadow regions are above minimum specified size. The third algorithm uses a threshold on the intensity reduction as a result of light being blocked by moving objects. This reduction depends on the background geometry, light source, moving object size, and position. The threshold reduction caused by a shadow is obtained by manually specifying shadow regions and recording these values which are later used for shadow removal in the same scene.