Improved depth propagation for 2D to 3D video conversion using key-frames

Overview of efficient high-quality state-of-the-art depth enhancement methods by thorough design space exploration

“…For key-frame propagation Varekamp et al [42] compute a temporal prediction of the next depth image from the previous one with a joint bilateral propagation filter (JP) given by…”

Superpixel Based Depth Map Generation for Stereoscopic Video Conversion

“…Bilateral filter which combines domain filtering and range filtering is very effective for depth map propagation [6]. In [6], the geometric distance and the color difference are used to determine the Gaussian weights in bilateral filter.…”

“…In [6], the geometric distance and the color difference are used to determine the Gaussian weights in bilateral filter. This can't distinguish different objects with similar color values.…”

“…The study in [5] first over-segments and manually annotates the original image, then detects the edge and T-junction, and finally obtains the depth map by depth propagation and post-processing. Varekamp et al [6] proposed a depth propagation method based on bilateral filtering through a blockbased motion compensation algorithm. Wu et al [7] use bidirection optical flow and Mean Shift algorithm to extract foreground object and track depth information for non-key frames.…”

“…Then, for non-key frames, the depth maps are generated automatically by a disparity propagation algorithm. The approaches in [5], [6] mainly aim at image semi-automatic depth generation. While approaches in [7]- [10] mainly focus on the video semi-automatic depth propagation based on the depth maps which have already been generated for the key frames with computational complexity.…”