Page 1 of 11
Journal for Studies in Management and Planning
Available at
http://edupediapublications.org/journals/index.php/JSMaP/
e-ISSN: 2395-0463
Volume 02 Issue 12
December 2016
Available online: http://edupediapublications.org/journals/index.php/JSMaP/ P a g e | 38
Region of exceptional detection by fusion of compactness
based on local prevalence and contrast
Ganuthala Upendar Reddy,2. Mr.P.V.Vara Prasad Rao
1PG Scholar, Department of ECE, SLC's Institute of Engineering and Technology, Piglipur Village, Hayathnagar Mandal, Near
Ramoji Film City, Ranga Reddy District, Hyderabad, Telangana
2Assosciate Professor, Department of ECE, SLC's Institute of Engineering and Technology, Piglipur Village, Hayathnagar
Mandal, Near Ramoji Film City, Ranga Reddy District, Hyderabad, Telangana
Abstract—Salient region detection is a
challenging problem and an important topic
in computer vision. It has a wide range of
applications, such as object recognition and
segmentation. Many approaches have been
proposed to detect salient regions using
different visual cues, such as compactness,
uniqueness, and objectness. However, each
visual cue-based method has its own
limitations. After analyzing the advantages
and limitations of different visual cues, we
found that compactness and local contrast
are complementary to each other. In
addition, local contrast can very effectively
recover incorrectly suppressed salient
regions using compactness cues. Motivated
by this, we propose a bottom-up salient
region detection method that integrates
compactness and local contrast cues.
Furthermore, to produce a pixel-accurate
saliency map that more uniformly covers the
salient objects, we propagate the saliency
information using a diffusion process. Our
experimental results on four benchmark
data sets demonstrate the effectiveness of the
proposed method. Our method produces
more accurate saliency maps with better
precision-recall curve and higher F- Measure than other 19 state-of-the-arts
approaches on ASD, CSSD, and ECSSD
data sets.
I. INTRODUCTION
Visual attention is an important mechanism
of the human visual system. It filters out
redundant visual information and effectively
selects highly relevant subjects, which are
called the salient objects. Visual attention is
considered to involve two mechanisms:
stimulus driven [1] and task driven . The
stimulus-driven mechanism is often called
bottom-up, and is fast, involuntary, and
purely based low-level visual stimuli. The
task-driven mechanism is called top-down,
and is based on high-level information such
as prior knowledge of the task, emotions,
and expectations. Accordingly,
Page 2 of 11
Journal for Studies in Management and Planning
Available at
http://edupediapublications.org/journals/index.php/JSMaP/
e-ISSN: 2395-0463
Volume 02 Issue 12
December 2016
Available online: http://edupediapublications.org/journals/index.php/JSMaP/ P a g e | 39
computational visual attention methods can
be categorized into bottom-up and top-down
methods. In this paper, we focus on bottom- up salient region detection tasks. Salient
region detection methods aim to completely
highlight entire objects of interest and
sufficiently suppress background regions.
Their output can be used for numerous
computer vision problems such as image
classification object detection and
recognition image compression and image
segmentation As a fundamental computer
vision task, salient region detection has been
extensively studied over the past few years,
and a number of algorithms have been
proposed Most bottom-up salient region
detection methods rely on visual cues to
consistently separate the salient object and
background. These cues include uniqueness
compactness and background.
Most uniqueness-based methods use low- level features of the image (such as
intensity, color, and orientation) to
determine the contrast between image pixels
or regions and their surroundings. According
to the contrastive reference regions, these
methods can be roughly divided into local- and global contrast-based methods. Local
contrast-based methods consider the
uniqueness of pixels (or superpixels, image
regions) with respect to their surrounding
regions or local neighborhoods, whereas
global contrast-based methods consider
contrast relationships over the entire image.
Unlike uniqueness-based methods, which
consider the uniqueness of low-level
features in the feature space, compactness
based methods consider the spatial variance
of features. Ideally, salient pixels (or
superpixels, image regions) tend to have a
small spatial variance in the image space,
whereas the background is distributed over
the entire image and has a high spatial
variance. Background based methods use
boundary and connectivity priors derived
from common backgrounds in natural
images . These methods are primarily
motivated by the psychophysical
observations that salient objects seldom
touch the image boundary, and most
background regions can be easily connected
to each other. Although the above- mentioned methods have achieved good
results in some aspects, each method has its
own limita-
Page 3 of 11
Journal for Studies in Management and Planning
Available at
http://edupediapublications.org/journals/index.php/JSMaP/
e-ISSN: 2395-0463
Volume 02 Issue 12
December 2016
Available online: http://edupediapublications.org/journals/index.php/JSMaP/ P a g e | 40
Fig. 1. Visual limitations of different
methods. (a) Input image. (b) Ground truth
salient regions. (c) Saliency maps using
local contrast based method . (d) Saliency
maps using global contrast based method
[7]. (e) Saliency maps using compactness
based method . (f) Saliency maps using
background based method . (g) Our method.
tions.
For example, Fig. 1 illustrates the saliency
detection results using four state-of-the-art
methods Figure 1(c) shows that the local
contrast-based method tends to highlight the
salient object’s edges instead of uniformly
propagating the saliency to the interior. The
global contrast based method sometimes
produces high saliency values for non- salient regions, especially for regions with
complex patterns or rare background
distractors. This is shown in the first
example of Fig. 1(d), where some grass
regions in the background are highlighted. A
typical limitation of the compactness based
method is that some salient regions may be
wrongly suppressed when the foreground
objects and background are similar. In the
second example of Fig. 1(e), the inner
smooth parts of the clock are wrongly
suppressed. Finally, background based
methods can perform very well. However,
they fail when the salient objects touch the
image boundary, as illustrated in the last two
examples of Fig. 1(f). From the above
discussion, we can conclude that single
visual cue based salient region detection
methods all have their own limitations. To
determine these limitations, different visual
cues should be integrated into a unified
framework. Motivated by this approach,
some methods integrate multiple visual cues.
Perazzi et al. [9] proposed a saliency filters
method, which unifies uniqueness and
compactness (of the spatial distribution) into
a single, high-dimensional, Gaussian
filtering framework. However, global
contrast and compactness based methods
have difficulty distinguishing between
similar colors in the foreground and
background. Consequently, the saliency
filters method fails when foreground objects
and the background are similar (e.g., the
second example in Fig. 1(e)). In this work,
we integrated local contrast and
compactness visual cues to generate saliency
E:\nitya\bhanu sir\Design and Operation of Wi-Fi Agribot Integrated System\f11.png
