A survey on coding binary visual features extracted from video sequences

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395 -0056

Volume: 04 Issue: 01 | Jan -2017

p-ISSN: 2395-0072

www.irjet.net

A SURVEY ON CODING BINARY VISUAL FEATURES EXTRACTED FROM VIDEO SEQUENCES Sreejaya1, Anu Vijayan2 , Athira Krishnan3 , Dhanya Sreedharan4 B.Tech Student, Department Of Computer Science and Engineering, Sree Buddha College Of Engineering, Alappuzha, Kerala, India 2 B.Tech Student, Department Of Computer Science and Engineering, Sree Buddha College Of Engineering, Alappuzha, Kerala, India 3 B.Tech Student, Department Of Computer Science and Engineering, Sree Buddha College Of Engineering, Alappuzha, Kerala, India 4Assistant Professor, Department Of Computer Science and Engineering, Sree Buddha College Of Engineering, Alappuzha, Kerala, India 1

---------------------------------------------------------------------***--------------------------------------------------------------------image and what humans recall after having observed

Abstract -

In pattern recognition and in image

an image or

processing, feature extraction starts from an initial set of

a

group local

of images after features

some

measured data and builds derived values intended to be

minutes.Binary

represent

an

informative and non-redundant, facilitating the subsequent

alternative to real-valued descriptors.A compact

learning and generalization steps, and in some cases leading

representation based on global features is preferred

to better human interpretations.When the input data to

when dealing with large collections.

an algorithm is too large to be processed and it is suspected

The visual content is acquired at a node,

to be redundant then it can be transformed into a reduced

compressed and then sent to a central unit for

set of features . Visual descriptors are descriptions of the

further processing according to the compress-then-

visual features of the contents in images, videos, or

analyze (CTA) paradigm in the case of traditional

algorithms or applications that produce such descriptions.

approach.. In the traditionally adopted compress-

KeyWords: Pattern recognition ,Visual features, Image

then-analyze (CTA) paradigm, images acquired from

processing, Feature extraction,Visual descriptors.

camera nodes are JPEG compressed and sent to a central controller for further analysis.In the case of

1.INTRODUCTION

analyze-then-compress

Feature extraction is a type of dimensionality

features and the relative keypoints information to a

approach is useful when image sizes are large and a

central controller. At the central controller, the

reduced feature representation is required toquickly

received features are matched against a database of

complete tasks such as image matching and

labeled features, so that object recognition or image

retrieval.Descriptors are the first step to find out the

retrieval can be performed .

connection between pixels contained in a digital Impact Factor value: 5.181

camera

transmit a compressed version of the extracted

of an image as a compact feature vector. This

|

approach

nodes perform visual features extraction.It then

reduction that efficiently represents interesting parts

Š 2017, IRJET

(ATC)

|

ISO 9001:2008 Certified Journal

| Page 1520

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395 -0056

Volume: 04 Issue: 01 | Jan -2017

p-ISSN: 2395-0072

www.irjet.net

2 LITERATURE SURVEY

2.3 Evaluation Of Low – Complexity Visual Feature Detectors And Descriptors

2.1 Coding Binary Local Features Extracted From Video Sequences

Feature detection selects regions of an image that have unique content, such as corners or blobs. Use

Local features refer to a pattern or distinct structure

feature detection to find points of interest that you

found in an image, such as a point, edge, or small

can use for further processing. These points do not

image patch. They are usually associated with an

necessarily correspond to physical structures, such

image patch that differs from its immediate

as the corners of a table. The key to feature detection

surroundings by texture, color, or intensity. Feature

is to find features that remain locally invariant so

extraction involves computing a descriptor, which is

that you can detect them even in the presence of

typically done on regions centered around detected

rotation or scale change.Binary descriptors can

features. Descriptors rely on image processing to

achieve a performance similar to that of non

transform a local pixel neighborhood into a compact

binarydescriptors,with

vector representation. This new representation permits

comparison

between

task as well as standalone.a good score in the

regardless of changes in scale or orientation.

standalone evaluation did not always lead to high

Descriptors, such as SIFT or SURF, rely on local

accuracy in the image retrieval task.

gradient computations. Binary descriptors, such as BRISK or FREAK, rely on pairs of local intensity

2.4 Rate -Accuracy Optimization Of Binary

differences, which are then encoded into a binary

Descriptors

vector.

Then-

Scale-invariant

Compress In

the

case

of

compress-then-analyze

lower

complexity.Evaluation is done for an image retrieval

neighborhoods

2.2 Compress-Then-Analyze VS. Analyze-

much

feature

transform (SIFT)

is

algorithm to detect and describe local features in

(CTA)

images. SIFT can robustly identify objects even

paradigm, images acquired from camera nodes are

among clutter and under partial occlusion, because

JPEG compressed and sent to a central controller for

the SIFT feature descriptor is invariant to uniform

further analysis.In the case of analyze-then-compress

scaling, orientation, and partially invariant to affine

(ATC) approach camera nodes perform visual

distortion and illumination changes.The descriptor is

features extraction.It then transmit a compressed

represented by means of a binary string, in which

version of the extracted features and the relative

each bit is the result of the pair-wise comparison of

keypoints information to a central controller. At the

smoothed pixel values properly selected in a patch

central controller, the received features are matched

around each keypoint.

against a database of labeled features, so that object recognition or image retrieval can be performed . Š 2017, IRJET

|

Impact Factor value: 5.181

an

|

ISO 9001:2008 Certified Journal

| Page 1521

International Research Journal of Engineering and Technology (IRJET)

e-ISSN: 2395 -0056

Volume: 04 Issue: 01 | Jan -2017

p-ISSN: 2395-0072

www.irjet.net

3 CONCLUSION

[4] Rate-accuracy optimization of binary descriptors

This paper

- A. Redondi, L. Baroffio, J. Ascenso, M. Cesano, and M.

tailored

deals with various coding schemes

to

both

local

and

global

binary

Tagliasacchi

features,which aims at exploiting both spatial and

[5] Video Google: A text retrieval approach to object

temporal redundancy by means of intra- and inter-

matching in videos - J. Sivic and A. Zisserman

frame coding.Binary local features represent an

[6] Distinctive image features from scale-invariant

alternative to real-valued descriptors.The efficiency

keypoints - D. G. Lowe

was evaluated by means of rate-efficiency curves

[7] Aggregating local descriptors into a compact

with respect to traditional visual analysis tasks. This

image representation - H. Jegou, M. Douze, C. Schmid,

survey is basically done to study about various

and P. Perez

coding schemes for extracting local and global binary

[8] Binary local descriptors based on robust hashing

visual features from video sequences.

- L. Baroffio, M. Cesana, A. Redondi, and M. Tagliasacchi

ACKNOWLEGEMENT

[9] BRIEF: Binary robust independent elementary

We are deeply thankful to our guide Ms Dhanya

features - M. Calonder, V. Lepetit, C. Strecha, and P.

Sreedharan, Assistant professor of Computer Science

Fua

and Engineering, for guiding us through the difficult

[10] BRISK: Binary robust invariant scalable

phases of our work and inspiring us during each

keypoints - S. Leutenegger, M. Chli, and R. Y. Siegwart

stage of our work. We express sincere thanks to all other faculty members for encouraging us in each stage.

REFERENCES [1] Coding binary local features extracted from video sequences - L. Baroffio, J. Ascenso, M. Cesana, A. Redondi, and M. Tagliasacchi [2] Compress - then - analyze vs. analyse - then compress: Two paradigms for image analysis in visual sensor networks - A. Redondi, L. Baroffio, M. Cesana, and M. Tagliasacchi [3] Evaluation of low - complexity visual feature detectors and Descriptors - A. Canclini, M. Cesana, A. Redondi, M. Tagliasacchi, J. Ascenso, and R. Cilla

Š 2017, IRJET

|

Impact Factor value: 5.181

|

ISO 9001:2008 Certified Journal

| Page 1522