Research on Depth Information Estimation for Video Objects

Abstract

Through the analyzing of the color, motion, parallax and the degree of clarity in scene-all the elements presented on the 2D video, People can produce 3D video smoothly. The paper introduces a kind of pick-up algorithm of depth map, the method of combining the abstraction of motion in depth based on the block matching and the abstraction of background in depth on the basis of background subtraction. The extracted map can be applied to the video conversion technology of "2D to 3D".

Simulating Shadow Interactions for Outdoor Augmented Reality With RGBD Data

“…Therefore, when handling shadow interactions between virtual objects and real objects, Jacob et al [7] and Jiddi et al [8] exploited the known 3D geometry of the real scene and the position of the light source to compute the shadow areas and simulated shadow interactions between virtual objects and real scenes. Although the 3D geometry that is provided by a depth camera or a vision-based algorithm such as depth estimation [9]- [11] or SLAM [12]- [14] can facilitate shadow interaction recovery, the main limitation of utilizing the 3D geometry of the scene to simulate shadow interaction is as follows: shadows may be visible even if the shadow-casting objects are not fully visible in the videos. For example, the right building in Fig.…”