基于FTV的改进深度估计算法

深度估计是任意视点电视(Free View Television，FTV)系统的关键技术。为提高深 度边缘的深度估计准确度，并解决视频序列中不同帧间同一深度物体深度估计值不一致问题 ，提出一种在深度图的时间一致性保持算法，以前一帧的深度图与当前视图的运动信息为约 束条件估计当前帧的深度，使得相邻帧的深度图在时间上保持一致。

An improved depth estimation algorithm based on FTV

Depth estimation is one of the key technologies of FTV(Free View Television) sys tem. To incr ease depth accuracy, a depth estimation method which can preserve temporal consi stency is proposed. For current view, depth values of the previous frame and mot ion information between current frame and its previous frame are used as constra int co nditions to estimate depth map of current frame, thus preserving temporal consis te ncy of depth maps in the two successive frames. Experiment results show that not only accuracy of depth edges in depth maps is increased, but also temporal cons istency of depth maps is preserved well.