PROTECTION OF 3D OBJECTS AGAINST ILLEGAL

PHOTOGRAPHY USING OPTICAL WATERMARKING

TECHNIQUE WITH SPATIALLY MODULATED

ILLUMINATION

Yasunori Ishikawa, Kazutake Uehira and Kazuhisa Yanaka

Kanagawa Institute of Technology, Atsugi-shi, Japan

Keywords: Watermarking, 3D objects, Spatially modulated illumination.

Abstract: We present a new technique that protects the copyrights or portrait rights of 3D objects such as sculptures,

merchandise, and even human bodies, with optical watermarking, which is produced by spatially modulated

illumination. Although the previous study revealed that the optical watermarking technique could prevent

objects from being illegally photographed without protection, the technique could only be applied to 2D

objects. The largest problem to be solved in extending this technique to the case of 3D objects is to

compensate for geometrical distortion. We solved this problem by introducing rectangular mesh fitting and

a technique of "bi-linear interpolation" based on the four nearest points. We conducted experiments in

which we projected optical watermarking onto the surface of a globe and a model of the human face, and

evaluated the accuracy of extracted data. The results were almost 100% in both cases when a Discrete

Cosine Transform (DCT) and a Walsh-Hadamard Transform (WHT) were used as methods of embedding

watermarks.

1 INTRODUCTION

Techniques of digital watermarking have been

widely recognized in recent years as methods of

protecting the copyrights of digital image content.

For example, digital watermarking is embedded in

digital data before it is printed to protect paper

images. This is the same as for other types of digital

media. However, this method cannot prevent

photographs of valuable paintings at museums and

galleries from being illegally taken with digital

cameras.

We have proposed a novel technology that can

prevent the illegal use of images of objects that do

not have watermarking, using illumination with

invisible watermarking. We carried out experiments

and revealed that watermarking data were read out

with almost 100% accuracy. In this paper we

propose the optical watermarking technology

applying to real 3D objects like sculptures in

museums, merchandise in stores, or even human

bodies on stages.

Figure 1: Basic concept underlying proposed technology.

Ishikawa Y., Uehira K. and Yanaka K..

PROTECTION OF 3D OBJECTS AGAINST ILLEGAL PHOTOGRAPHY USING OPTICAL WATERMARKING TECHNIQUE WITH SPATIALLY

MODULATED ILLUMINATION.

DOI: 10.5220/0003314400490052

In Proceedings of the International Conference on Imaging Theory and Applications and International Conference on Information Visualization Theory

and Applications (IMAGAPP-2011), pages 49-52

ISBN: 978-989-8425-46-1

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

2 WATERMARKING

TECHNIQUE APPLYING

TO BASIC 2D OBJECTS

Fig. 1 outlines the basic concept underlying the

optical watermarking technology. A projector is

possibly used as the light source that contains the

watermarking information and illuminates an object.

The brightness of the object's surface is proportional

to the product of the reflectance of the surface and

the illumination by the light source.

Fig. 2 illustrates the procedure for watermarking

using orthogonal transforms. The watermarking area

is divided into units of 16×16 or 8×8 pixel blocks,

and each block has a DC component that gives an

average brightness for the entire watermarking area,

i.e., the brightness of illumination. Every block also

has the highest frequency component (HC) in both

the

- and

- directions to express the 1-bit binary

information for watermarking. The phase of HC was

used to express binary data i.e., "0" or "1". Other

components than DC or HC were set to "0". DCT

and WHT were used as orthogonal transforms to

produce the watermarking images. The equations of

i-DCT and i-WHT are expressed with Eq. (1) and Eq.

(2) respectively.

Figure 2: Producing watermarks.







),()()(),(

vuFvCuCyxf

(1)

}

)12(

cos{}

)12(

cos{





(, ) (,) (,) (, )

ij ij

xy F uvwhxuwhvy







(2)

where

),(

yxf

are the watermarking image data for

pixel

),( yx

of block

),( ji

in real space,

),(

vuF

are

the data for component

),( vu

of block

),( ji

frequency space, and

is the number of pixels in

the block in the

- and

-directions. Here,

)(uC

and

)(vC

are given as

)0(2

)0(1

)(





)0(2

)0(1

)(





and

),( jiwh

denotes a component of the Walsh-

Hadamard matrix in Table 1.

Table 1: Walsh-Hadamard Matrix.

(a) 8×8 Matrix

(b) 16×16 Matrix

3 APPLYING TO 3D OBJECT

First, the grid pattern image that divides the region

of the optical watermarking equally into 8 × 8 is

irradiated onto the object image and captured with a

digital camera to apply optical watermarking

technique to 3D objects. The coordinates of the

corner points of the segmented areas are measured

IMAGAPP 2011 - International Conference on Imaging Theory and Applications

respectively on the captured image data. Then, the

image data irradiated with optical watermarking are

captured, and these are corrected by using the

coordinates of the corner points of the segmented

area as each segmented region may become a precise

square.

The transformation from an undistorted coordinate

system

),( yx

to a geometrically distorted system

)','( yx

is generally expressed by following

equations.

),('

yxhx 

),('

yxhy



(3)

If the distortion is perspective, the transformation is

expressed by the following linear equations.

cbyaxx '

feydxy 



(4)

Using these equations the value of pixels in the

distorted quadrangle can be transformed to the value

of pixels in the undistorted rectangle. However,

because the coordinates of transformed pixels do not

generally become integers, an interpolation

technique is utilized to determine the density value

of the nearest pixel. Linear transformation using the

four nearest neighboring pixels was used in the

experiments, which is so called "bi-linear

interpolation".

4 EXPERIMENTS

Watermarking images had 128 × 128 pixels that

consisted of 16 × 16 blocks of 8 × 8 pixels in the

experiments. A Digital Light Processing (DLP)

projector that had a resolution of 800 × 600 pixels

was used as a light source. A white hemisphere, a

globe, and a model of a human face were used as

real 3D objects. Fig. 3. (a) shows the image of a

globe irradiated with the grid pattern, and Fig. 3. (b)

and (c) have the photographs of the human-face

model onto which the watermarking or grid pattern

was projected.

Using the measured coordinates of the grid

points, each 8 × 8 divided segments were identified

and were corrected to a precise rectangle as

described in the previous section. The restored

rectangular area had about 650 × 650 pixels using a

digital camera with a resolution of 4288 × 2848

pixels. It was transformed to 256 × 256 pixels and

divided into 16 × 16 blocks. We carried out DCT on

all blocks using Eq. (5).











),(

)()(

),(

jiji

yxf

vCuC

vuF

}

)12(

cos{}

)12(

cos{





(5)

We also utilized Eq. (6) for WHT, using the values

in Table 1(b) as the components of matrix

),( jiwh

and



(,) (, ) (,) (,)

ij ij

uv f xywhuxwhyv







(6)

The accuracy of reading out the embedded data was

evaluated by checking the sign of the

)7,7(

, ji

components for all blocks. Two methods of

embedding data were used. The "1-block method"

involved embedding 1-bit data into one block and

embedding 256 1-bit binary data into 16 × 16 blocks.

The "majority method" involved embedding the

same 1-bit data into three blocks sufficiently

separated from one another, and the readout data

were determined by the majority decision. The

distance between the blocks was set to five and 75 1-

bit binary data were embedded in 16 × 16 blocks in

the experiments.

(a) Grid pattern on a globe (b) Human-face model (c) Magnified image of the grid

pattern on a human-face model

Figure 3: Photographs of a globe and a human-face model on which image was projected.

PROTECTION OF 3D OBJECTS AGAINST ILLEGAL PHOTOGRAPHY USING OPTICAL WATERMARKING

TECHNIQUE WITH SPATIALLY MODULATED ILLUMINATION

Table 2: Experimental results: Accuracy with which embedded data were read out.

HC →

5 7 10 15 20 25 5 7 10 15 20 25

white-ball 111111111111

globe-1 0.964 0.964 0.964 0.966 0.982 0.996 110.987 111

globe-2 0.996 0.996 1111111111

face-1 1 0.996 0.991 111111111

face-2 111111111111

white-ball 111111111111

globe-1 0.969 0.987 0.991 0.982 0.996 1111111

globe-2 111111111111

face-1 111111111111

face-2 111111111111

WHT

1-block method Majority method

DCT

5 RESULTS AND DISCUSSION

Table 2 lists the overall results of experiments. In the

table "white-ball" means that a white hemisphere

was used as 3D object. In the same way, "globe-1"

means European-African hemisphere of a globe was

used, "globe-2" means Pacific-Ocean hemisphere of

a globe was used, "face-1" means cheek of a human-

face model was used and "face-2" means forehead of

a human-face model was used. The results of

evaluating accuracy for the white hemisphere had

100% under all conditions with DCT and WHT.

However, 100% accuracy was not achieved in

evaluating accuracy with the globe, especially in the

European-African hemisphere, where the 1-bit block

method was used with DCT and WHT. The decision

by using the majority method achieved an accuracy

of 100% excluding the HC=10 of DCT. The

European-African hemisphere has numerous black

lines and characters and these could disturb the

accuracy of reading out embedded data.

An accuracy of 100% for the evaluation of the

human-face model was obtained under all conditions

in the decision by the majority method. The 1- block

method achieved an accuracy of 100% excluding

part of the DCT. The surface of the human-face

model was painted white in this experiment, and the

reflectivity of the surface may have been

proportional to the brightness of the irradiated

optical watermarking.

6 CONCLUSIONS

We proposed the application of optical watermarking

to 3D objects, which can prevent real objects like

sculptures in museums from being illegally

photographed. We used methods of correcting

distortions in captured images caused by projecting

optical watermarking image onto the curved surface

of objects. We found that the embedded data were

read out with almost 100% accuracy when DCT and

WHT were used for embedding watermarking, after

distortions in the captured images had been

corrected. In this paper we used a projected grid

pattern to indicate the correct pixel block, prior to

the images for embedding watermarking being

captured. However, if the marker that appropriately

identifies pixel blocks is simultaneously embedded

into optical watermarking images, it can easily be

extracted with image processing. Therefore, we

demonstrated the feasibility of using optical

watermarking technique to protect real 3D objects

from being illegally captured which has been

difficult to accomplish with conventional

watermarking technology.

REFERENCES

Mizumoto, T. and Matsui, K. 2002. Robustness

investigation of DCT digital watermark for printing

and scanning. Trans. IEICE (A), J85-A(4), 451-459.

Uehira, K. and Suzuki, K. 2008. Digital watermarking

technique using brightness-modulated light. Proc.

ICME2008, 257-260.

Ishikawa, Y., Uehira, K. and Yanaka, K. 2009.

Illumination watermarking technique using orthogonal

transforms. Proc. IAS2009, 257-260.

Kishk, S. and Javidi, B. 2003. 3D object watermarking by

a 3D hidden object. OPTICS EXPRESS, 11(8), 874-

888.

Bennour, J. and Dugelay, J. L. 2006. Protection of 3D

object through silhouette watermarking. Proc. ICASSP

2006.

Rosenfeld, A. and Kak, A. C. 1979. Digital Picture

Processing (pp. 175-179). New York: Academic Press.

IMAGAPP 2011 - International Conference on Imaging Theory and Applications