Visual pertinent 2D-to-3D video conversion by multi-cue fusion

Abstract

We describe an approach to2D-to-3D video conversion for the stereoscopic display. Targeting the problem of synthesizing the frames of a virtual "right view" from the original monocular 2D video, we generate the stereoscopic video in steps as following. (1) A 2.5D depth map is first estimated in a multi-cue fusion manner by leveraging motion cues and photometric cues in video frames with a depth prior of spatial and temporal smoothness. (2)The depth map is converted to a disparity map with considering both the displaying device size and human"s stereoscopic visual perception constraints. (3)We fix the original 2D frames as the "left view" ones, and warp them to "virtually viewed" right ones according to the predicted disparity value. The main contribution of this method is to combine motion and photometric cues together to estimate depth map.In the experiments, we apply our method to converting several movie clips of well-known films into stereoscopic 3D video and get good results 1 .

A Compact Representation for Compressing Converted Stereo Videos

“…Fig. 5(a) displays another scene, whose complex background depth model is estimated by an automatic algorithm [34] and the foreground object is represented with a simple planar surface model. These results show the flexibility of the proposed CVCVC system in estimating scene depth.…”

“…The TransFantasy system automatically converts 2D videos into 3D. In [34], motion and aerial perspective cues are leveraged to estimate depth maps. In [33], disparity maps, but not depth maps, are directly estimated by a trained SVM using features of object motion and context motion.…”

A Compact Stereoscopic Video Representation for 3D Video Generation and Coding

“…At the decoder side, a 2 rom bit-stream, and then the depth cues ar ding to image features of 2D video. At last, the n a DIBR method [9].…”

“…I lossless 2D shape coding alg to select a minimum numbe from a dense object contou shape can still be recovere adopt the Intelligent Scissors to skip redundant contour p features as shape recovery cu the selected control points ar the object contour can be f using image features as algorithm is described in Alg lattice is represented by a g weight on a graph edge is def 3: An example of background modeling em, near objects are labeled in detail, but for matic depth estimation (may not be so accurate) ithm proposed in [9] to automatically estimate t ws:…”

“…For background, if 0 in , background depth map is estimated by the automatic depth estimation algorithm [9] from its corresponding 2D video frame; otherwise, the decoded stage model is adopted. For foreground, object contours are recovered by applying Intelligent Scissors on a 2D frame between each pair of neighboring control points .…”