Database Reference
In-Depth Information
Fig. 7.5
Top five retrievals (
b
)-(
f
) answering to a query clip that contains two video shots (
a
),
using the group-to-group querying method. Note that the first ranked video is the query itself. All
of the retrieved clips are originally from the same story
This section introduces a novel way to implement Graph Cut for video object
segmentation with shape information. Graph Cut is a very efficient algorithm for
image segmentation, whereas Histogram of Oriented Gradients (HOG) is useful
in detecting humans. The HOG feature is combined to incorporate a shape prior
into the Graph Cut algorithm as a new way to enhance video object segmentation
accuracy.
Graph Cut methods were applied for image and video segmentation. For image
segmentation [
192
], a 2-D graph is constructed from the image color information,
with each pixel representing a node in the graph. The nodes are connected by arcs or
edges that represent the energy cost for
cutting
that edge. This is the pairwise energy,
commonly assigned a cost relative to the intensity difference between adjacent pairs
of pixels. For video object segmentation, the structure of Graph Cut is conducted as
Search WWH ::
Custom Search