Occlusion handling in Augmented Reality context
Abstract: Handling occlusion between real and virtual objects is a challenging problem in Augmented Reality (AR) applications. Incorrect and inaccurate occlusion handling may cause confusion in users’ perception which leads to non-realistic and non-immersive AR experiences. Even though there are important research examples and implementations on this topic, they typically suffer from important limitations. The method presented in this thesis takes both raw depth and color information from an RGB-D sensor in order to solve occlusion in unknown AR contexts. The novelty of this method is the automatic trimap generation which is based on the color distribution information of the real scene. Moreover, the proposed method can operate using the information from only the latest frame, or efficiently use the information from multiple frames. In the second case, the knowledge of the previous frame output is used to improve the trimap generation of the latest frame. Experimental evaluations of several scenes demonstrate that this approach largely improves automatic trimap generation using both the RGB and depth information. Furthermore, the final proposed method is compared with some recent state of the art approaches in terms of quality and accuracy, showing that it overcomes some of the known limitations.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)