DIGITAL RECORDING OF TV BROADCASTING AND

ADVERTISEMENT DETECTION

Francisco Ortiz

Department of Physics, Systems Engineering and Signal Theory, University of Alicante, Spain

Alberto Gómez

Department of Engineering, University Miguel Hernández, Spain

Keywords: Advertisement detection, Artificial vision, Pattern recognition.

Abstract: In this paper we provide an integrated approach for detecting commercial advertisements in TV

broadcasting for a digital recording of old TV shows. This approach uses Artificial Vision for detecting the

TV logo in the video streams. The results of comprehensive experiments on a heterogeneous data emission

set of 300 minutes of video taken from 5 different sources are reported. Our method provides almost 98%

correct detection and elimination of the advertisements for a digital recording. We offer an effective

solution with the new software developed.

1 INTRODUCTION

The new video digital formats allow an optimization

in the storage of the TV shows of the TV enterprises.

An important detail to reduce the recording volume

is to eliminate the advertising of programs recorded

on older video formats

On the other hand, it is important to control the

emission of advertising in the media. In the focus of

the market studies and control of TV publicity, some

precedent of a digital system of advertising control

exists in (Campoy) present ADETECT (Automatic

Detection and Recognition of TV Advertisements)

as a program for the detection of announcements in

an advertising emission, comparing them with a data

base of announcements, that is, this system identifies

that announcement is emitting itself at every

moment. This detection is made comparing first

frames of the announcements with those of the data

base through areas or landmarks in the image, using

artificial vision concepts. This system for 4 years it

has been commercialized and implanted in several

consultancies of several countries to take the control

of which announcements are being emitted and to

that hour relay it.

A similarity system, for written press, has been

developed by Centre de Vision by Computer of the

University of Barcelona and in the program of

European Union EUTIST-IMV (Integrated Machine

for Vision European Industry), (INFOADEX). This

system facilitates the control of the publicity in mass

media.

We offer an effective solution to advertising

control and commercial-free digital recording by

means of visual identity in the live broadcast.

2 THE PROJECT:

DETECTION AND

ELIMINATION IN TV

BROADCASTING

The main of our system is the automatic

discrimination between advertisements and TV

shows o films in TV broadcastings. Artificial vision

will permit us to recognize emission sequences

continuity of programming or advertising emission.

Some similar systems use audio information

(Marlow, 2001). Our method is the detection of the

symbol or logo of the TV enterprise in the image.

The TV logo is appearing/disappearing

when there is

program emission or advertisements. TV disappears

in publicity.

595

Ortiz F. and Gómez A. (2010).

DIGITAL RECORDING OF TV BROADCASTING AND ADVERTISEMENT DETECTION.

In Proceedings of the 2nd International Conference on Agents and Artiﬁcial Intelligence - Artiﬁcial Intelligence, pages 595-598

DOI: 10.5220/0002726105950598

 SciTePress

The television channels or enterprises that have

been used in our study are: “1-tve”, “2-tve”,

“Antena3”, “Tele5” and “Canal9”.

The developed system has been applied to the

digital recording by a professional video from the

analogical broadcasting.

The program will detect, by the recognition of

logo, if the emission is valid or not, and it will come

to send a command to the digital video (REC or

STOP). A digital recording of program TV which is

free of advertising announcements is obtained.

A very interesting utility of our project is the

conversion of TV emission from analogical to digital

media, without advertisements or publicity. This is

very useful in TV producers in which they keep a

file from the emitted thing, since when keeping the

daily emission also stores the emitted publicity

(commercial control).

On the other hand, also it is useful this system to

have a control of time emission of advertisements in

TV enterprises. This information is very important

for TV channels and ANUNCIANTES

3 SYSTEM DESCRIPTION

Next, we detail the hardware and software

characteristics of our system for detecting and

eliminating advertisements in TV broadcasting:

 Digital video. JVC Digital-S BR-D750E, digital

video of professional quality used in this

investigation to have port series and

communicate command from the computer

with standard Sony protocols in professional

video.

 Analogical video. Standard analogical video

system.

 PC. Standard PC. Intel Pentium D. Windows

XP operating system.

 Digital card. “Matrox Corona”. This hardware

permits to digitize in PAL and NTSC system,

and composite video signal, according to norm

CCIR by Y/C and RGB. Also, it can capture

non-standard systems. True quality color 24

bits.

 Software: The original software has been

developed in C++ language.

The TV signal comes from the analogical

video and it goes to the digitizer "Matrox Corona"

in PC. This same signal is also derived to the

digital video that will record the information

coming from the analogical video. The signal

produced by the analogical video or analogical

antenna can be visualized through the screen of the

PC using the program "Intellicam" that provides

Matrox.

4 STEPS OF THE ALGORITHM

FOR DIGITAL RECORDING

WITHOUT TV

ADVERTISEMENTS

Our system for advertisement detection of TV

broadcasting consists of several modules that make

different functions.

One first previous task to the execution of the

algorithm is the save in a file of the patterns of TV

logos with which the study will be made. They will

be saved in binary format. Without these patterns,

we can not detect the valid emission (without TV

advertisements). We save together the patters a

threshold of coincidence of logos. This will help to

differentiate them to each other and from the

background in which they are located and to obtain a

greater success in the detection. In the following

table we show a percentage of coincidence between

TV logos. Also, we work with another superior

threshold that allows us to differentiate the TV logo

shape from the background. In our case, 80% have

been considered as optimal as recognition of the

shape of the logo.

In Figure 1, we can see a logo TV detection in a

corner of the original image of a TV sequence of

“Canal9”.

Once the system gets the logos from file, the

algorithm begins a “do…while” loop. Initially, four

small images corresponding to four corners of the

TV emission are acquired in where the TV logos are

located. The algorithm searches the TV logos in the

four corner images every 0,5 seconds. If there is not

similarity, there is a change in the process from REC

to STOP. If there is similarity with a TV logo, then,

the system is REC the analogical TV broadcasting to

a digital system. Also, in the next iteration of the

main loop, only the same corner is analyzed.

When the system has located the logo TV, it

shows the information by a control window: a logo

TV "v

" with a coincidence level "p

" is in the corner

“c

" and shows the image and its four processed sub-

images. Then, it sends the command "REC" to the

digital video for the recording the emission. If the

system does not search any logo TV, the algorithm

makes four consecutive verifications every 0.5

seconds without sending any command to the digital

video. The command “STOP” is sent to digital video

once these cycles have finished. This preserves to

the system of continuity errors of TV emission: in

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

596

Figure 1: Logo TV detection in a corner of the original image.

Figure 2: All the steps of the algorithm for detecting TV advertisements.

order to maximise the visual impact of TV

advertisements, producers frequently use a faster

rate of shot cuts than normal TV content

(Hauptmann). Figure 2 shows all the steps of the

algorithm for detecting TV advertisements.

In the algorithm, the “loop” is in process until the

user paralyzes the system. In any case, the command

“STOP” is sent to the digital video when this

happens.

BEGINNING (cont=0)

Data introduction

Data acquisition

End of loop?

if cont = 4

Send “STOP”

Creation of logo

patrons

The system finds the

o TV?

Show

information

Send “STOP”

Send “REC”

Show options

Show information

YES NO

if cont < 4

cont++

THE

END

DIGITAL RECORDING OF TV BROADCASTING AND ADVERTISEMENT DETECTION

597

5 RESULTS

Now, show a significant example of the results

obtained in the execution of our system of recording

without advertising (Table 1). This example

represents 5 hours of emission (300 minutes) 60

minutes by each one of the TV enterprise that

compose the study (percentage in erroneous state):

Table 1: Results of advertisement detection.

TV enterprise percentage in erroneous state

“1-tve” 2,2%

“2-tve” 1,83%

“Antena3” 2,1%

“Tele5” 3,4%

“Canal9” 2,8%

Of the 60 minutes of test by TV transmitter, it

has been arrived at a high efficiency, then in the case

of “2-tve”, only there is a difference of 35 seconds

(1,83% in erroneous state). In all the cases the

system detects without failure some logo when this

one is present in the image and the emission is valid.

The over-recording has been procured until being

completely safe of not finding logo of the transmitter

in the image.

6 CONCLUSIONS

We can conclude that the developed system fulfills

the expectations wished and it can to be used to

eliminate the advertisement in new digital recording

of old TV shows.

At present, we work on improving the algorithm

of digital recording. It aims to decrease the error rate

to less than 1%, by increasing the speed of detection

of the TV logo. We are reducing the computation

time of the algorithm, improving the “identification”

functions (pattern recognition).

In addition, we are making adjustments to the

system for recording digital TV, reducing, in this

case the complexity of the tasks.

REFERENCES

Campoy, P. ADETECT: Automatic Detection and

Recognition of TV Advertisements. In

http://www.disam.upm.es/vision/projects/adetect/adete

ct.pdf.

INFOADEX. Control y Análisis de la publicidad. In:

http://www.infoadex.es/infoadex.asp.

Marlow, S., Sadlier, D. et al. Audio and Video Processing

for Automatic TV advertisement Detection, In Proc.

Irish Signals and Systems Conference. 2001.

MATROX IMAGING LIBRARY (MIL), available in

http://www.matrox.com/imaging/products/mil/b_mil.p

Hauptmann, A., Witbrock, M. “Story Segmentation and

Detection of Commercials in Broadcast News Video”,

Proceedings of Advances in Digital Libraries

Conferenc., 1998.

ICAART 2010 - 2nd International Conference on Agents and Artificial Intelligence

598