TAG INTERFACE FOR PERVASIVE COMPUTING

Paper Tag Interface using Imae Code

Dong-Chul Kim, Jong-Hoon Seo

Media System Laboratory, Dept. of Computer Science, Yonsei University, Seoul, Korea

Cheolho Cheong, Tack-Don Han

Media System Laboratory, Dept. of Computer Science, Yonsei University, Seoul, Korea

Keywords: 2D barcode, Image code, Tag interface, Pervasive computing, Ubiquitous computing.

Abstract: Recently, computing environments move to pervasive computing age with rapid growth of internet and

appearance of various mobile devices. This means computer offers more convenient life with linking

between physical object and digital information. With these advance of computing environment, several

researches are progress that are about tag interface to link between various physical objects and digital

information, which are image code like as barcode, or wireless technology based like as RFID. It leads high

expenses to read or write because RFID has to buy tags – as much as wanted – and has to need exclusive

hardware. But image code can be printed to paper and can decode from camera connected to computer, so it

is convenient and less maintenance cost. In this paper, we developed encoding and decoding algorithm for

image code and applied algorithm to the tag interface to access the information more easy and fast in

pervasive computing environment and develop.

1 INTRODUCTION

As the paradigm of computing evolves to pervasive

computing, physical objects and digital information

are being linked organically and life goes more

convenient. These changing affect not only

performance, but also user usability to more human-

friendly manner. By these evolution, it is more

important in pervasive computing environment that

linking naturally between physical space and digital

space with using tag interface. Tag interface

connects between digital information of virtual

space – which is in the computing environment or on

the network - and real objects in the physical world.

Many researches are in advance to progress

computing environment and to make more

intelligent with tag interface. There are so many

input devices - e.g. keyboard, mouse, character

recognition, etc - which digitalize physical

information. However, some drawbacks are exist,

e.g. many user effort, slow input speed, and high

error rate. Also, to digitalize huge information which

has complicated structure takes so many time and

frequent errors. [Kambayashi, 2001]. To overcome

these drawbacks, we proposed tag interface. Tag

interface is classified into RFID and image based

code. Image based code is used in various fields as

like barcode. Many services based on tag interface

use RFID instead of barcode. Because RFID tags

become more miniaturized, high effectiveness, and

less-expenses. However to construct these system, it

needs exclusive tags, readers, and writers, so initial

investment expenses are increased. Also, it is

difficult that user approach, and at cognition process,

many drawbacks exist - i.e. interferences, distance

recognition.

In this paper, we proposed paper tag interface

which uses image-based code to overcome

drawbacks of conventional RFID limited user

access, and expensive cost, and to diversify

application fields. At section 2, we survey tag

interface related conventional works, and section 3

shows proposed paper tag interface. At section 4, we

apply variety practical fields, and section 5

concludes the paper.

2 IMAGE CODE

Conventional image-based codes are divided into 1D

bar-code, and 2D image code. Firstly, 1D code is the

barcode that can be met in general. However, 2D

image codes are classified into Black-and-White

codes which use just two colors, and colourful code

356

Kim D., Seo J., Cheong C. and Han T. (2006).

TAG INTERFACE FOR PERVASIVE COMPUTING - Paper Tag Interface using Imae Code.

In Proceedings of the International Conference on Signal Processing and Multimedia Applications, pages 356-359

DOI: 10.5220/0001572003560359

 SciTePress

- this code uses multiple colors to enhance the

capacity of the code in Table 1. The 1D bar-code is

used in several industries, with merit about fast

recognition and direction-free cognition ability. It is

most popular tag interface which is used in

circulation, logistics, etc. Also it has very high

recognition rate about 100%, and it can be printed

various place, i.e. paper, wrapping vinyl, etc, and

with inexpensive cost, and it can be diverse size. It

consists of white background and black lines which

has different thickness, and is recognized using CCD

or laser. Code system for barcode is supported in

EAN (European Article Number) which is

international standard and UPC code system.

However it has a defect that the size of the code

would be larger to represent a large size of data,

since the code has limited capacity and even lower

recording density. In 2D image codes, developed to

overcome a defect of 1D barcode - i.e. low capacity

and density, they have been developed widely, but

they have to be used with expensive and exclusive

readers only.

Table 1: Generation of image based code.

First Second Third

Code 1D Barcode 2D Barcode

2D Color Image

code

Keyword

Speed,

Correctiness

Capacity,

Speed, Error

Correction

Color, Mobility

Example

EAN/UPC,

Codabar, etc

QR code,

PDF417,

Data Matrix,

Maxi code, etc

ColorCode,

Ultracode, etc

Many codes have been developed for large

capacity and density. Denso have made QR code

which can be recognized quickly. Symbol

Technologies developed PDF 417 that has high

restoration rate[Pavlidis, 1992]. International Data

Matrix made the Data Matrix which can be made in

small size. UPS developed the Maxi Code to

separate postal matter in speedy. Sony made the

CyberCode for augmented reality. The Ultracode

from Zebra, the DataGlyphs from Xerox[Hecht,

2001], the ColorCode from ColorZip Media and so

many codes are developed, and these codes have

several advantages - e.g. data capability, data type

can be represented, recording density, etc - than 1D

barcode. Of these, the ColorCode and the Ultracode

use multiple colors and can store more data in same

size[Palmer, 1995]. In these semantics, 2D barcode

has portable data file and can be the data bridge

between computers they do not connected each

other[Want, 1999]. In other words, a data file printed

in 2D symbology from a computer system can re-

input to another system without keyboard.

3 PAPER TAG INTERFACE

SYSTEM ARCHITECTURE

Paper tag interface system plays music when user

puts down a music tag on the guide line in front of

computer. When paper tag is put, the system

processes recognition algorithm from the inputted

camera image. That algorithm composed of

Binarization image, analysis component, and

detection candidate area of code. We decode image

code and recognize data if no error occurs from

candidate code area. We explain the structure of

image code and recognition process in 3.1. The

computer offers the user a service by performing

action – e.g. playing music, open internet browser,

etc – after tag recognition. In case of music tag, the

computer serves user through one of output devices

like as a speaker. We proposed two types of tags.

The one is card type and that’s size is 8.5 x 5.3 (cm)

in Fig. 1. Another is label type size of 5.5 x 1.5 (cm)

in Fig. 2. Label type is more portable than card type.

In case of card type, it has icon image in centre, Fig.

1-(1). The icon can help user recognize the purpose

of the tag. Bottom of the tag – Fig. 1-(2), there exist

image code for computer cognition and there is

additional information for user in Fig. 1-(3) – e.g.

the title, the singer, etc.

Figure 1: Paper Tag Structure of Card Type.

In case of label type, it has advantages of

portable and keeping. There is image code in Fig. 2-

(1) and additional information for user in Fig. 2-(2).

There are no icon image areas, because this type is

so small.

Figure 2: Paper Tag Structure of Label Type.

(1) Icon image area

(2) Image Code

(3) Additional Information

(1) Image Code

(2) Additional Information

TAG INTERFACE FOR PERVASIVE COMPUTING - Paper Tag Interface using Imae Code

357

User can generate tags easy as printing on paper.

Paper tag can contain some process of any action

and optional data or any file. If the data size is over

than image code capacity, then the data can be

uploaded server and image code has server address

and information of authentication.

3.1 Designing Image Code

We design image code for paper tag interface.

Proposed image code is designed by circular strap

Fig. 3-(1). The centre of code has directional

indicator to inform the decoder the starting position

of this code, as like Fig. 3-(2). The circular strap is

formed data bits toward clockwise. Left 45 degree

and right 45 degree diagonal means '0' bit and '1' bit.

Reading data sequence ends at last data bits of k

layer.

Figure 3: Imae Code Structure.

If we let the index of circle at the code be 1, 2, …,

k from the most inner circle to the most outer circle,

the total capacity of data in the k

circle is

proportional to the circle index k. Let r

be the radius

of k

circle and n

be the number of codes in k

circle. Then we can say r

krr

(1)

And, we say θ

is a central angle of fan-shape

which a single pattern occupied. As the size of each

pattern is same, each pattern has uniform area and

all of patterns have fixed length of circumference.

(2)

(1) is substituted for (2), then

krrl

(3)

Therefore the number of codes in the k

layer is,

(4)

As we can see above, we designed to proportion

the total number of patterns in k

circle with circle

index k.

3.2 Detecting Code Area and

Recognition Algorithm

The algorithm – for detecting code area and

recognition the code – can be briefly described as

following. At first, decoder pre-processes for

separating code area from the background image at

the inputted image. Filtering process filters

candidate areas. This process reduces operation

loads of decoding. After detection of code area,

decoder computes centre point. Decoder reads

circular data strap from the start bit that was found

by directional indicator, and recognize all of data

sequence after parity check. More detailed

description is explained as following 4 steps.

Step 1: Binarization

Decoder thresholds inputted original image, Fig.

4- (1), to binarized image, Fig. 4-(2) before

analysis image. Firstly, to find the value of

threshold, decoder converts original image to grey,

and calculates an average value from min-max

histogram. This process called “dynamic

binarization”, which provides more flexible

adaptation to various environment than fixed value

thresholding.

Step 2: Compnent Shape Analysis

Firstly, to get shape information as like Fig. 4-(3),

we apply laplacian mask to binarized image and get

contour which have connected component

information by tracking edge lines. By applying

laplacian mask, we can get good results and less

amount of computation loads to get edge. Especially,

it is easy to track contour information because

laplacian masking produces 1pixel edge result from

the binarized image.

Step 3: Candidate Filtering

We get the area which is estimated as code field,

with following two filters from the previous contour.

The first filter is the outline length information. We

use the outline information to remove the noise area.

If the length of the outline is too short, we would

estimate it as small object of background or noise, so

it would be estimated as the background. The second

filter is circular rate. We filtered out non-code areas

from the result of first filter with using circular rate.

We define circular rate e as following equation.

/4 lAre ×=

(5)

Above equation (5) is gained from the circular

area formula (

) and circular circumference

formula (

2 rl

). The more the circular rate e is

near to 1.0, the more it can be estimated as a circular

Data

Strap(3)

Directional

Indicator(2)

Circular

Strap(1)

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

358

shape and it would be decides as the final candidate

code area, Fig. 4-(4). We can decrease the

calculation quantity by filtering non-code area out.

Step 4: Decoding and Error Detection

We compute the center point and the radius from

the extracted candidate area, and look for the pattern

around the center point circularly. Firstly, we search

the indicating sign which is showing the start bit. To

start to read the data, we find the start bit from the

most outer layer in the direction of the indicating

sign. A bit pattern means 0 or 1 – inclined left(‘/’) is

0 and inclined right(‘/’) is 1. After decoding patterns

we apply error check algorithm to detect errors. If

there are no errors detected, it can be recognized as

correct codes, and then decoding process is regarded

as done.

(1) Original Image (2) Binarization

(3) Edge Extraction (4) Get Image code Area

Figure 4: Recognition through Preprocessing.

4 EXPERIMENTAL RESULTS

AND APPLICATION

We organized 4 layers for test, and data is inserted

with ratio of 2:3:4 from the 2

layer to 4

layer. We

tested the printed picture of the 2.5 × 2.5 (cm) size

code. We could save the data of 147bits in the used

test code if it except the parity code. The size of the

data would increase exponentially when more layers

appended. At above size, maximum layer can be

recognized are 7 and then more than 800 bits can be

contained. We experiment paper tag system as

following example. We applied the tags as a paper

storage media, security and verification keys,

multimedia educations system for children,

computer interface of easy approach, entertainments

interface (photo, music, move) for a media center

computer.

(1) File Case Tag (2) Paper Storage

(3) Music Tag (4) E-mail Tag

Figure 5: Paper Tag Applications.

5 CONCLUSION

In this paper, we have made up tag interface using

image code system architecture for easy accessing to

digital information in pervasive computing

environments. We design and apply image code in

this system. It can be possible that there are no needs

to hardware devices like RFID. User can use card

type or label type tag for their purpose. It is easier

and cheaper than RFID tag system to generate tags

by printing on paper and to decode by using PC

camera. We develop algorithm of encoding and

decoding for image code generation and recognition.

The following points are left as future problems.

Generation of paper tag with no use computer

system and increasing tag capacity.

REFERENCES

Kambayashi, Y., Tarumi, H., Morishita, K., 2001. Digital

Tags: Data with Restricted Accessibility for e

Commerce Applications, Australian Computer Science

Communications.

Pavlidis, T., Swartz, J., and Wang, Y. P., 1992.

Information Encoding with Two-Dimensional Bar

Codes, IEEE Computer Magazine.

Hecht, D.L., 2001. Printed embedded data graphical user

interfaces, IEEE Digital Object Identifier.

Palmer, R. C. 1995. The Bar Code Book, third edition,

Helmers Publishing.

Want, R., Fishkin, K.P., Gujar, A., Harrison, B.L., 1999.

Bridging Physical and Virtual Worlds with Electronic

Tags, Proc. of CHI’99.

TAG INTERFACE FOR PERVASIVE COMPUTING - Paper Tag Interface using Imae Code

359