Review On Different Feature Extraction Algorithms

© 2016, IRJET | Impact Factor value: 4.45 | ISO 9001:2008 Certified Journal | Page 58
Review On Different Feature Extraction Algorithms
Shilpa G. Sonone1, Dr. Mahadev Kokate2
Department of Electronics and Telecommunication Engineering
K.K.Wagh Institute of Engineering Education & Research
Nashik 422003, INDIA
email:
1shilpasonone04@gmail.com,
2mdkokote@kkwagh.edu.in
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract -: In visual sensor networks, energy
conservation is main issue because more energy is
consumed in processing and transmitting the large
amount of data. The traditional approach, denoted here
in after as “Compress-Then-Analyze” (CTA), consists in
the following steps: the visual content is acquired by a
sensor node in the form of still images or video
sequences; then, it is encoded and efficiently
transmitted to a central unit where visual feature
extraction and analysis takes place. In this process
large amount of energy is consumed in compression
processes and transmission processes also transmitting
data at low bandwidth is also a main issue so to
overcome this issues new approach is used i.e. “Analyze-
Then-Compress” (ATC) In particular; nodes process
visual content in order to extract relevant information
in the form of visual features. Then, such information is
compressed and sent to a central unit, where visual
analysis takes place. Here In this paper we just
overview the different compression techniques or
different algorithms in real valued (SIFT And SURF) and
binary feature extraction (BRIEF, BRISK, Bin Boost) and
also describe coding technique in WVSN.
Key Words: Visual features, binary descriptors, SIFT,
SURF, BRIEF, BRISK, Bin-Boost, video coding.
1. INTRODUCTION
Visual features provide a succinct, yet efficient,
representation of the underlying visual content, which is
robust and invariant to many global and local
transformations. They are effectively employed in many
tasks, ranging from image/video retrieval, object
recognition, object tracking, image registration, structure-
from
motion, etc.[1]Visual feature extraction algorithms consist
of two main components: the detector, which identifies
salient keypoints within an image; and the descriptor,
which provides a concise representation of the image
patch surrounding each keypoint. There are different
descriptor are designed till now which perform same
processes such as namely pre-smoothing, transformation
and spatial pooling.[2]
In WVSN different compression techniques are used to
compressed data and transfer it to the base station. The
tradition approach is “Compress-Then-Analyze”(CTA)is
based on the following steps: the signal of interest (i.e., a
still image or a video sequence) is acquired by a sensor
node, then it is compressed(e.g., resorting to JPEG or
H.264/AVC coding standards) in order to be efficiently
transmitted over a network. Finally, visual analysis is
performed at a sink node [3]– [6].This type is used in
different application like video surveillance, smart
cameras, etc. but it represent a lossy visual features and
having bandwidth limitation so we switch toward new
technique Analyze-Then-Compress.
An alternative technique to the Compressed-then-Analyze
is Analyze –then-Compressed is gaining popularity in the
research community. In this process local features are
extracted directly from sensing node. Then, they are
compressed to be efficiently dispatched over the network.
There are different algorithms are used for extraction of
local features such as SIFT, SURF, BIREF, BRISK, Bin-Boost
etc.
2. RELATED WORK
Compression of visual features is one of the
important task in WVSN. The main problem related to
transmission of date is bandwidth, in WVSN the date
which is transferred to the sink node having low
bandwidth. Several works tackled this problem for the
case of features extracted from still images, proposing
methods to efficiently encode state-of-the-art visual
features [7]or to modify the design of local feature
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 03 Issue: 02 | Feb-2016 www.irjet.net p-ISSN: 2395-0072

extraction algorithms, so that the representation of the
underlying visual content is more suitable for
compression [8]. In this context, an ad-hoc MPEG group on
Compact Descriptors for Visual Search (CDVS) is currently
working towards the definition of a standard tailored to
this scenario [9].
Processing of the visual features require more
data so the energy require for transmission of the large is
also more so this is also one of the problem regarding
transmission of data directly to the base station[10]. So
different algorithms are used to extract visual features and
then designed encoder which encodes the data at low
computational complexity.
First, Real-valued features such as SIFT [11] or
SURF [12],These are methods for extracting distinctive
invariant features from images that can be used to
perform reliable matching between different views of an
object or scene. The features are invariant to image scale
and rotation distortion. This also describes an approach to
using these features for object recognition. The
recognition proceeds by matching individual features to a
database of features from known objects using a fast
nearest-neighbour algorithm, followed by a Hough
transform to identify clusters belonging to a single object,
and finally performing verification through least squares
solution for consistent pose parameters. This approach to
recognition can robustly identify objects among clutter
and occlusion while achieving near real-time performance.
BRIEF [13] propose to use binary strings as an
efficient feature pointed scriptor, which we call BRIEF. We
show that it is highly discriminative even when using
relatively few bits and can be computed using simple
intensity difference tests. Furthermore, the descriptor
similarity can be evaluated using the Hamming distance,
which is very efficient to compute. Thus this algorithm is
suitable whenever energy resources area issue, such as in
the case of low-power devices, where they constitute the
only available option. The processing pipeline for the
extraction of local features comprises: i) a keypoint
detector, which is responsible for the identification of a set
of salient keypoints within an image, and ii) a keypoint
descriptor, which assigns a description vector to each
identified keypoint, based on the local image content.
Within the class of local binary descriptors, BRIEF
[14]computes the descriptor elements as the result of
pairwise comparisons between (smoothed) pixel intensity
values thatare randomly sampled from the neighbourhood
of a keypoint.
BRISK [14], FREAK [15] and ORB [16] are
inspired by BRIEF, and similarly to their predecessor, are
also based on pair wise pixel intensity comparisons. They
differ from each other in the way pixel pairs are spatially
sampled in the image patch surrounding a given keypoint.
BAMBOO [17] adopts a richer dictionary of pixel
intensity comparisons, and selects the most discriminative
ones by means of a boosting algorithm. This leads to a
matching accuracy similar to SIFT, while being 50x faster
to compute. Bin Boost [18],which proposes a boosted
binary descriptor based on a set of local gradients. Bin
Boost is shown to deliver state-of-the-art matching
accuracy, at the cost of a computational complexity.
In this paper we just overview the different
algorithms of extracting features or compare different
types of descriptors for video coding. Now in next part we
explain proposed system of video coder by using BRISK
algorithm, then proposed flow chart of video coder and
BRISK algorithm, and finally conclusion.
3. PRAPOSED SYSTEM OF VIDEO CODER
As described in previous section different compression
techniques are used in WVSN and different algorithms are
used for visual feature extraction. Here we describe video
coder in which features extracted from video by using
binary feature extraction algorithm i.e. BRISK algorithm.
After extraction of features by BRISK algorithm then it
coded by using entropy coding as shown in figure 1.then
this data is transferred wireless channel to the base
station. Here descriptors are coded to meet the energy and
bandwidth requirement
Local Feature
Detection
Descriptor
Extraction
Entropy
Coding
Camera
Node
Wireless
Channel
Descriptor
Decoding
Descriptor
Matching
Image
data
Query Result
Fig.1 Proposed Block Diagram for video coding and
feature matching.

Figure1. describe the proposed block diagram of video
coding and feature matching. In this first Image is capture
from camera node and then keypoints are detected. then
descriptors are extracted from image then these points are
encoded by using BRISK algorithm here Descriptors
keypoints are coded in binary form. Then this data is
encoded by using entropy coding and transferred to the
base station by using wireless channels.
At decoder side i.e. at base station this data is first decoded
by using descriptor decoder then it given to the descriptor
matching. Here in descriptor matching comparison is done
between decoded descriptor’s data and Image database
which is already stored at base station so query result is
obtained here decoder side is used for video retrieval task
in WVSN.
In this system we used Binary Feature extraction
algorithm i.e. BRISK algorithm is used for feature
extraction. This Brisk algorithm is more suitable than
other algorithm like SIFT and SURF. The flow chart of
Brisk algorithm is given in Figure 2. Here Figure 2a shows
flowchart of keypoint detection and Figure 2b shows
Flowchart of descriptor designing.
Fig.2a Flow chart of Keypoint detection
Separate pairs of pixels into
two subsets
Compute local gradient
Sum gradients to determine
feature orientation
Rotate short-distance pairs
using orientation
Rotate short-distance pairs
using orientation
Sampling Pattern
Construct binary descriptor
from rotated short-distance
pairs
Stop
a
Fig 2b.Flow chart of Descriptor Designing
4. CONCLUSIONS
We conclude that Analysis-then-Compression is much
better than Compressed-then-Analyzed method. And also
conclude that by comparing different algorithms of
compression that binary feature extraction algorithms are
more suitable than SIFT and SURF algorithm.
ACKNOWLEDGEMENT
The authors can acknowledge any person/authorities in
this section. This is not mandatory.
REFERENCES
[1] L. Baroffio, M. Cesana, A. Redondi, M. Tagliasacchi,
and S. Tubaro, “Coding visual features extracted
from video sequences,” IEEE Trans.Image Process.,
vol. 23, no. 5, pp. 2262–2276, May 2014.
[2] L. Baroffio, M. Cesana, A. Redondi, S. Tubaro, and
M. Tagliasacchi. (2014, Mar.). Coding Visual
Features Extracted from Video. Dept. di Elettron.,
Informazionee Bioingegneria, Politecnicodi
Milano, Milan, Italy, Tech. Rep. DEIB-2014-1

[Online].Available:http://home.deib.polimi.it/tagl
iasa/greeneyes/technote.pdf
[3] B. Tavli, K. Bicakci, R. Zilan, and J. Barcelo-Ordinas,
“A survey of visual sensor network platforms,”
Multimedia Tools Appl., vol. 60, no. 3,pp. 689–726,
2012.
[4] K. Obraczka, R. Manduchi, and J. J. Garcia-Luna-
Aveces, “Managing the information flow in visual
sensor networks,” in Proc. 5th Int.Symp. Wireless
Pers. Multimedia Commun., vol. 3. Honolulu, HI,
USA, Oct. 2002, pp. 1177–1181.
[5] S. Lee, S. Lee, and A. Bovik, “Optimal image
transmission over visual sensor networks,” in
Proc. 18th IEEE Int. Conf. Image Process., Brussels,
Belgium, Sep. 2011, pp. 161–164.
[6] A. Canclini, L. Baroffio, M. Cesana, A. Redondi, and
M. Tagliasacchi, “Comparison of two paradigms
for image analysis in visual sensor networks,” in
Proc. 11th ACM Conf. Embedded Netw. Sensor Syst.,
New York, NY, USA, 2013, pp. 62:1–62:2.
[7] V. Chandrasekhar, G. Takacs, D. Chen, S. S. Tsai, J.
Singh, and B. Girod, “Transform coding of image
feature descriptors,” in VisualCommunications and
Image Processing, vol. 7257, M. Rabbani and R. L.
Stevenson, Eds. Bellingham, WA, USA: SPIE, 2009,
pp. 725–710.
[8] V. Chandrasekhar et al., “Compressed histogram
of gradients: A lowbitrate descriptor,” Int. J.
Comput. Vis., vol. 96, no. 3, pp. 384–399, Feb. 2012.
[9] MPEG, Denver, DC, USA. Compact Descriptors for
Visual Search [Online]. Available:
http://mpeg.chiariglione.org/standards/mpeg-
7/compact-descriptors-visual-search
[10] Luca Baroffio, Antonio Canclini, MatteoCesana,
Alessandro Redondi, Marco Tagliasacchi, and
Stefano Tubaro,” Coding Local and Global Binary
Visual Features Extracted From Video Sequences”,
IEEE transactions on image processing, VOL. 24,
NO. 11, pp1-15,november 2015
[11] D. G. Lowe, “Distinctive image features from
scale-invariant keypoints,” Int. J. Comput. Vis., vol.
60, no. 2, pp. 91–110, Nov. 2004.
[12] H. Bay, T. Tuytelaars, and L. Van Gool, “SURF:
Speeded up robustfeatures,” in Proc. ECCV, 2006,
pp. 404–417.
[13] M. Calonder, V. Lepetit, C. Strecha, and P. Fua,
“BRIEF: Binary robust independent elementary
features,” in Proc. ECCV, 2010, pp. 778–792.
[14] S. Leutenegger, M. Chli, and R. Y. Siegwart, “BRISK:
Binary robust invariant scalable keypoints,” in
Proc. ICCV, Nov. 2011, pp. 2548–2555.
[15] A. Alahi, R. Ortiz, and P. Vandergheynst, “Freak:
Fast retina keypoint,” in Proc. CVPR, Jun. 2012, pp.
510–517.
[16] E. Rublee, V. Rabaud, K. Konolige, and G. Bradski,
“ORB: An efficient alternative to SIFT or SURF,” in
Proc. IEEE Int. Conf. Comput.Vis. (ICCV), Nov. 2011,
pp. 2564–2571.
[17] L. Baroffio, M. Cesana, A. Redondi, and M.
Tagliasacchi, “Bamboo: A fast descriptor based on
AsymMetric pairwise BOOsting,” in Proc.IEEE Int.
Conf. Image Process. (ICIP), Oct. 2014, pp. 5686–
5690.
[18] T. Trzcinski, M. Christoudias, V. Lepetit, and P.
Fua, “Boosting binary keypoint descriptors,” in
Proc. IEEE Conf. Comput. Vis. PatternRecognit., Jun.
2013, pp. 2874–2881.

Review On Different Feature Extraction Algorithms

More Related Content

What's hot (20)

Similar to Review On Different Feature Extraction Algorithms (20)

More from IRJET Journal (20)

Recently uploaded (20)

Review On Different Feature Extraction Algorithms