Flash shadow detection and removal in stereo photography

Shadows created by flash when taking stereo images adversely affect 3D perception when stereo images are merged into 3D images. In this study, a shadow detection and removal algorithms are proposed for stereo images taken under flash lighting. Foreground object is extracted from left and right images. Shadow seed area is set right next to the foreground object in order to decide the pixel value range of the shadow. After choosing shadow candidate, originally dark area is removed. After the detection, shadow area is brightened up according to the reference area. This is the first attempt at addressing the flash shadow problem in stereo photography. A quantitative measure is introduced to evaluate the effectiveness of shadow detection and a subjective rating is done to evaluate the effectiveness of shadow relighting or removal for 3D perception. It is shown that the developed solution outperforms the previous solutions designed for single images.