Digital Signal Processing Reference
In-Depth Information
Fig. 1.10
An example of
focus saliency map.
Left
:
Original image
gandalf
.
Right
: The focus saliency
map
Fig. 1.11
An example of image pairs for co-segmentation.
Left
: Image pair
amira
.
Right
:Image
pair
stone
an image can be computed from the difference between the original and the blurred
images. Most of the energy in the saliency map corresponds to the focused object,
whilst a large amount of the energy of the defocused region is removed efficiently.
In the second stage, bilateral and morphological filtering are employed to smooth
and accentuate the salient regions. The third stage involves adaptive error control
matting scheme to extract the boundaries of the focused objects. An example of
focus saliency map is shown in Fig.
1.10
, where most human regions have higher
saliency values than background.
1.3.4
Towards Object Driven Segmentation
A typical work of object driven segmentation is called “co-segmentation”, which
aims to segment the similar object from images. The main idea of co-segmentation
is to discover common objects from image pairs based on the assumption that each
image contains the foregrounds with similar color, texture, or shape. The first row of
Fig.
1.11
shows two image pairs with similar objects in the foreground but different
