Content Based Fingerprinting for Video Copyright Detection System

(1)

International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459, Volume 2, Issue 8, August 2012)

240

Content Based Fingerprinting for Video Copyright Detection

System

Diana Andrushia.A1, Shilpa Harikumar2

Assistant Professor, Department of ECE, Karunya University, Coimbatore, TamilNadu

PG Student, Department of ECE, ,Karunya University, Coimbatore, TamilNadu

Abstract — Security of a fingerprinting system is

important for certain applications such as copyright protections, for this reason secure version of the fingerprinting system for video piracy detection is presented. The proposed system relies on a fingerprint extraction algorithm followed by a fast approximate search algorithm for video piracy detection. The fingerprint extraction algorithm extracts content based signatures from special images constructed from the video. Each such image represents a short segment of the video and contains temporal as well as spatial information about the video segment. These images are denoted by temporally informative representative images. To find whether a query video is copied from a video in a video database, the fingerprints of all the videos in the database are extracted and stored in advance. The search algorithm searches the stored fingerprints to find close enough matches for the fingerprints of the query video. The proposed algorithm is designed and realized using MATLAB

Keywords— Content-based fingerprinting, video copy

detection, video copy retrieval

I. INTRODUCTION

Lots of videos are being uploaded to internet and shared every day. These videos may be manipulated versions of existing video, making copyright management on the internet a difficult process. Since videos are available in different formats, so it is better to base the copy detection process on the content of the video rather than its name. A fingerprint is a content-based signature derived from a video that represents the video. To find a copy of a query video in a video database, the close match of its fingerprint in the corresponding fingerprint database is obtained from the videos in the database. Closeness of two fingerprints represents a similarity between the corresponding videos and two different videos will be having different fingerprints. This work,deals with spatio-temporal fingerprints because of their comprehensiveness.

The temporally informative representative images (TIRIs) are formed from a video sequence since TIRI contains both spatial and temporal information of a short segment of a video sequence.Two fast search methods are used to matches the piracy of the video.

II. RELATED WORKS

As video is the most complex type of digital media, it has so far received the least attention regarding copyright management There has been great deal of work in the area of video fingerprinting .Some spatio-temporal algorithms consider a video as a three-dimensional (3-D) matrix and extract 3-D transform-based features [7], [2]. Others use spatio-temporal interest-point descriptors to generate the fingerprints [5], [3]. Ordinal spatio-temporal features have also been used in the literature [4]. Applying a 3-D transform to a video is a computationally demanding process and may pose problems in online applications. In [5] the proposed method for forming temporally informative representative images (TIRIs) from a video sequence. Because videos are available in different formats, it is more efficient to base the copy detection process on the content of the video rather than its name, The fingerprints should also be discriminant, to ensure that two perceptually different videos have distinguishable fingerprints. The TIRI-DCT method along with the search algorithm introduces a fingerprinting system that is robust, discriminant and proves to be the best in providing accurate results when compared to other fingerprinting techniques.In the proposed video copyright detection system consists of the following stages

(2)

International Journal of Emerging Technology and Advanced Engineering

241 The basic block diagram of the fingerprinting system is given below. After obtaining the fingerprint feature matching is done. The fingerprint that matches with the fingerprint in the database can be declared as a pirated version .

[image:2.612.75.254.211.339.2]

Figure 1:Block diagram of fingerprinting system

III. GENERATION OF SPATIO TEMPORAL INFORMATION

USING TIRI-DCT

Figure 2 Block diagram of binary fingerprint Extraction

The input video is converted to frames and change the input size to 128*128 ie the width and height of the frame and divide the input into short segments of the video sequence.After which the binary fingerprints are extracted using the TIRI fingerprinting algorithm described below.

A.TIRI DCT Algorithm

Step 1

:

Generate the TIRI from each segment of J frames using wk=αk

Step 2: Segment each TIRI into overlapping blocks of size 2w*2w as follows:

Bi,j={l1x.y ǀ x € iw+w,y € jw+w

Step3: Extract two DCT coefficients from each blocks.These are the first horizontal and vertical coefficients adjacent to the DCT coefficients.

The first vertical frequency can be found out for Bi,j as follows

αi,j =VT Bi,j I

where V=cos{ 0.5/2w}….cos{∏-0.5 ∏ / 2w}

Step 4:Concatenate all the coefficients to form the feature vector f.

Step 5:Find m which is the median of all elements of f.

Step 6:Generate the binary hash h from f using the formula hk={1,if fk >= 1

0,if fk <= m

This method calculates a weighted average of the frames to generate a representative image. The resulting image is an image that contains information about possible existing motions in a video sequence. The TIRI is thus generated as follows: let lm,n,k be the luminance value of the

th pixel of the k th frame in a set of J frames. The pixels of TIRI are then obtained as a weighted sum of the frames

.

L1

m,n

=∑w

k*

l

m,n,k.

w

k represents the weight of eack k frames in J frames

[image:2.612.331.547.383.577.2]

B. Schematic of TIRI DCT algorithm

Figure 3: TIRI DCT implementation.

IV. SEARCHING TECHNIQUES

A.Inverted-File-Based Similarity Search

(3)

International Journal of Emerging Technology and Advanced Engineering

242 0010001 0010111………0010101

First word Second word nth word.

[image:3.612.100.255.211.363.2]

Figure 4: Position of a word i fingerprint.

Figure 5: Inverted file for the fingerprint database

To find a query fingerprint in the database, first the fingerprint is divided into words (of bits). The query is then compared to all the fingerprints that start with the same word. The indices of these fingerprints are found from the corresponding entry in the first column of the inverted file table. The Euclidean distance between these fingerprints and the query is calculated. If a fingerprint has a Euclidean distance of less than some predefined threshold which is the median of the spatio temporal features, it will be declared as the match. If no such match is found, the procedure is repeated for the fingerprints that have exactly the same second word as the query’s second word.This procedure is repeated until the last word is checked for. the algorithm is guaranteed to find the correct match, if th < n and if th>n the algorithm gives a false negatives. Probabilty of false rejection is

pFR=∑(-1)k(n)((L-k*m)/L)

B. Cluster-Based Similarity Search

By assigning each fingerprint to one and only one cluster. the fingerprints in the database will be clustered into groups. To determine if a query fingerprint matches a fingerprint in the database, the cluster head closest to the query is found. All the fingerprints (of the videos in the database) belonging to this cluster are then searched to find a match, i.e., the one which has the minimum Euclidean distance from the query. If a match is not found, the cluster that is the second closest to the query is examined.

[image:3.612.355.540.249.449.2]

This process continues until a match is found or the cluster far away is found out. In the second case no exact match is found. if the query is not matched to any fingerprint in a certain cluster, the algorithm continues its search over other clusters starting from the closer ones. For this method,all the fingerprints will be searched. The cluster heads should be chosen such that a small change in the fingerprint does not result in the fingerprint being assigned to another cluster.

Figure 6: Clustering of the fingerprints

Figure 7: Expanding a cluster head to compare it with a fingerprint in the database.

V. PERFOMANCE EVALUATION

[image:3.612.366.525.485.576.2]

(4)

International Journal of Emerging Technology and Advanced Engineering

243 Based on the binary fingerprint exrtracted from the query video, able to compare it with the fingerprint in the database using searching techniques to determine wheather a video is a pirated version of the video in the database. The thresholds were set to 0.26 for TIRI-DCT.In the proposed method distance measure performs much better than other distance for duplicate detection for shorter queries. This distance computation procedure, also end up computing which model vector serves as the best match for a query vector.This helps in faster identification of the best matching model keyframe for a given query keyframe.The results are obtained for binary fingerprint extraction using TIRI DCT fingerprinting algorithm and two searching techniques are used for fingerprint matching namely inverted file similiarity search and cluster based similiarity search.By calculating true positive rate and false rejection rate the performance are measured.

VI. RESULTS

[image:4.612.50.293.344.563.2]

Frame rate:10 frames/sec Duration of video :12sec

[image:4.612.361.526.474.591.2]

Figure. 8:.Query video

Figure.9:video database having different videos with different duration

Table 1

Table 2

Table 1 and Table2: Euclidean distances computed for each database videos for inverted file and cluster search methods

A.PERFORMANCE ANALYSIS

Figure 10 Accuracy vs search time for each searching technique

Input Euclidean

distance

1.car1 2.color led 3.fdsfds 4.ground2 5.ground3 6.hockeyground 7.video

1.6649 1.584 1.6263 1.7539 1.6263 1.8861 1.0121

Input Euclidean distance 1.car1

2.color led 3.fdsfds 4.ground2 5.ground3 6.hockeyground 7.video

(5)

International Journal of Emerging Technology and Advanced Engineering

244

INFERENCE

Using True positive rate accuracy of 84.6154% is obtained for Cluster-Based Similarity Search and 70.2308 % for Inverted-File-Based Similarity Search in terms of the search speed with which query is matched with the database videos.

VII. CONCLUSION

Fingerprinting algorithm for video copyright detection system has been proposed. It can be used for copyright management and indexing applications. TIRI-DCT algorithm extracts robust, discriminant, and compact fingerprints from videos in a fast and reliable fashion. Two fast approximate search algorithms: the inverted-file-based method and cluster based method has been implemented and the results have been compared based upon the true positive rate. The cluster-based method is superior to inverted-file-based method in terms of the detection performance.This work can be extended by taking large number of videos in a database with longer time duration.

REFERENCES

[1 ] Esmaeili.M.M‖A Robust and Fast Video Copy Detection System

Using Content-Based Fingerprinting‖ IEEE transaction on

Information security, volume 6, pp 213-226, march 2011

[2 ] J. Law-To, L. Chen, A. Joly, I. Laptev, O. Buisson, V.

Gouet-Brunet, N. Boujemaa, and F. Stentiford, ―Video copy detection: A comparative study,‖ in Proc. ACM Int. Conf. Image and Video Retrieval, New York, NY, 2007, pp. 371–378, ACM.

[3 ] A. Hampapur and R. M. Bolle, Videogrep: Video copy detection

using inverted file indices IBM Research Division Center, 2008.

[4 ] L. Chen and F. W. M. Stentiford, ―Video sequence matching based

on temporal ordinal measurement,‖ Pattern Recogn. Lett., vol. 29, no. 13, pp. 1824–1831, 2008.

[5 ] R. Radhakrishnan and C. Bauer, ―Content-based video signatures

based on projections of difference images,‖ in Proc. MMSP, Oct. 2007, pp. 341–344.

[6 ] C. De Roover, C. De Vleeschouwer, F. Lefebvre, and B. Macq,

―Robust video hashing based on radial projections of key frames,‖ IEEE Trans. Signal Process., vol. 53, no. 10, pp. 4020–4037, Oct. 2006.

[7 ] S. Lee and C. Yoo, ―Robust video fingerprinting for content-based video identification,‖ IEEE Trans. Circuits Syst. Video Technol., vol. 18, no. 7, pp. 983–988, Jul. 2005.

[8 ] X. Su, T. Huang, and W. Gao, ―Robust video fingerprinting based on

visual attention regions,‖ in Proc. ICASSP,Washington, DC, 2009, pp. 1525–1528, IEEE Computer Society.

[9 ] A. Joly, O. Buisson, and C. Frelicot, ―Content-based copy retrieval using distortion-based probabilistic similarity search,‖ IEEE Trans. Multimedia, vol. 9, no. 2, pp. 293–306, Feb. 2003.

[10 ]D. G. Lowe, ―Distinctive image features from scale-invariant

keypoints,‖ Int. J. Comput. Vis., vol. 60, no. 2, pp. 91–110, 2004.

[11 ]B. Coskun, B. Sankur, and N. Memon, ―Spatiotemporal transform

based video hashing,‖ IEEE Trans. Multimedia, vol. 8, no. 6, pp. 1190–1208, Dec. 2006.

[12 ]R. Radhakrishnan and C. Bauer, ―Robust video fingerprints based on

subspace embedding,‖ in Proc. ICASSP, Apr. 2008, pp. 2245–2248.

[13 ]G.Willems, T. Tuytelaars, and L. Van Gool, ―Spatio-temporal

features for robust content-based video copy detection,‖ in Proc. ACMInt. Conf. Multimedia Information Retrieval, New York, NY, 2008, pp. 283–290, ACM.

[14 ]A. Joly, O. Buisson, and C. Frelicot, ―Content-based copy retrieval using distortion-based probabilistic similarity search,‖ IEEE Trans. Multimedia, vol. 9, no. 2, pp. 293–306, Feb. 2007.

[15 ]C. Kim and B. Vasudev, ―Spatiotemporal sequence matching for