Given a video sequence, obtaining accurate layer segmentation
and alpha matting is very important for various applications.
However, when a non-textured or smooth area
is present in the scene, the segmentation based on only single
motion cue usually cannot provide satisfactory results.
Conversely, the most matting approaches require a smooth
assumption on foreground and background to obtain a good result.
In this paper, we combine the merits of motion segmentation and
alpha matting technique together to simultaneously achieve
high-quality layer segmentation and alpha mattes. First, we
explore a general occlusion constraint and design a novel graph
cuts framework to solve the layer-based motion segmentation
problem for the textured regions using multiple frames. Then, an
alpha matting technique is further used to refine the
segmentation and resolve the non-textured ambiguities by
determining proper alpha values for the foreground and
Jiangjian Xiao and Mubarak Shah, “Accurate
Motion Layer Segmentation and Matting”, IEEE Conference on Computer Vision and Pattern
Recognition, June 20-26, San Diego, 2005.
Jiangjian Xiao and Mubarak Shah, “Motion
Layer Extraction in the Presence of Occlusion using Graph Cut”,
IEEE transactions on Pattern Analysis and Machine Intelligence,
Vol. 27, No. 10, pp. 1644-1659, 2005
Figure 1: Multi-frame segmentation for a
two-layer synthetic image sequence. Top: several black lines,
two ellipses, and one triangle is moving form top-right to
bottom-left on a random fixed background. (a) The results
obtained using our previous algorithm with the occlusion order
constraint, where the pixels (red) occluded between frame 1 and
2 will be always occluded between frame 1 and j (j > 2). (b) The
correct results by our approach using frame 1 as the reference
image. (c) The correct results by our approach using frame 3 as
the reference image.
Figure 1: The results of flower-garden sequence. (a)
Original images. (b) Segmentation results before applying
matting process. (c) Alpha matting results. (d) Segmentation
results after matting process. (e) Object transfer after
the foreground of flower-garden in another video sequence.