OSA's Digital Library

Optics Express

Optics Express

  • Editor: Michael Duncan
  • Vol. 14, Iss. 4 — Feb. 20, 2006
  • pp: 1458–1467
« Show journal navigation

Fully-phase asymmetric-image verification system based on joint transform correlator

Hsuan T. Chang and Chao-C. Chen  »View Author Affiliations


Optics Express, Vol. 14, Issue 4, pp. 1458-1467 (2006)
http://dx.doi.org/10.1364/OE.14.001458


View Full Text Article

Acrobat PDF (449 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

A fully-phase optical asymmetric-image verification system based on a joint transform correlator (JTC) is proposed in this study. Conventional joint power spectra in JTCs are used as the amplitude information for reconstructing only the symmetric target images at the output plane. A previous method, in which an additional phase mask is used at the frequency domain as the phase information, proposed by Chang and Chen [9] was proposed to enable the reconstruction of asymmetric images at the output plane. However, the dominating effect arose from the additional phase makes the wrongly reconstructed image recognizable when the phase key at the input plane is incorrect. In the proposed method, the joint power spectra is nonlinearly transformed into the phase information for reconstructing both symmetric and asymmetric images at the output plane, while the dominating effect in the previous method can be released as well. Simulation results of using two different nonlinear transformations with different parameters are provided to verify the proposed method.

© 2006 Optical Society of America

Fig. 1. Optical setup of the original JTC architecture for symmetric image verification.

1 Introduction

Optical joint transform correlators (JTCs) [1

1. C.J. Weaver and J.W. Goodman, “A technique for optically convolving two functions,” Appl. Opt. 5, 1248–1249 (1966) [CrossRef] [PubMed]

] play an important role in optical signal processing techniques such as the image correlation framework in pattern recognition. Figure 1 shows the optical setup of a conventional JTC architecture. There are two input images h 1(x, y) and h 2(x, y) and their joint Fourier transform SI(u, v) is detected by the use of a charge-coupled device (CCD) in the intensity form (i.e., the joint power spectrum). By passing the joint power spectrum to the spatial light modulator (SLM) as the transmittance signal O(u, v), the cross-correlation and convolution signals of two input images can be obtained at the output plane.

Recently JTCs have shown the wide applications on optical security [2–5

2. B. Javidi and J.L. Horner, “Optical pattern recognition for validation and security verification,” Opt. Eng. 33, 1752–1756 (1994) [CrossRef]

]. Two major categories are the optics-based image encryption [3

3. B. Javidi and T. Nomura, “Polarization encoding for optical security systems,” Opt. Eng. 39, 2439–2443 (2000) [CrossRef]

] and security image verification [2

2. B. Javidi and J.L. Horner, “Optical pattern recognition for validation and security verification,” Opt. Eng. 33, 1752–1756 (1994) [CrossRef]

,4

4. T. Nomura and B. Javidi, “Optical encryption using a joint transform correlator architecture,” Opt. Eng. 39, 2031–2035 (2000) [CrossRef]

,7

7. J. Rosen, “Learning in correlators based on projection onto constraint sets,” Opt. Lett. 18, 1183–1185 (1993) [CrossRef] [PubMed]

,8

8. H.T. Chang and C.T. Chen, “Enhanced optical image verification based on joint transform correlator adopting Fourier hologram,” Opt. Rev. 11, 165–169 (2004) [CrossRef]

] systems. For the applications on image verification, there is no input plain image and both the input functions h 1(x,y) and h 2(x,y) at the input plane are phase-only That is, h 1(x,y) = exp[i2πp 1(x,y)] and h 2(x,y) = exp[i2πp 2(x,y)], where both functions p 1(x,y) and p 2(x,y) are the random numbers within the range [0, 1]. Given a specific target image o(x, y) at the output plane, both the phase functions in the input plane can be determined by use of some phase-retrieval algorithms [6

6. J.R. Fienup, “Phase retrieval algorithm: a comparison,” Appl. Opt. 22, 2758–2769 (1982) [CrossRef]

]. Initially, both phase functions are randomly generated. The phase function h 2(x,y) is fixed and refers to as a lock and the other phase function h 1(x,y) refers to as a key. To obtain a target image o(x,y) at the output plane for the fixed phase function h 2(x,y), the other one h 1(x,y) is iteratively retrieved using the phase retrieval algorithms such as the projection onto constraint set (POCS) algorithm [7

7. J. Rosen, “Learning in correlators based on projection onto constraint sets,” Opt. Lett. 18, 1183–1185 (1993) [CrossRef] [PubMed]

].

One of the major limitations of the conventional image verification systems based on JTCs is that only the symmetric images can be reconstructed at the output plane [5

5. D. Abookasis, O. Arazi, J. Rosen, and B. Javidi, “Security optical systems based on a joint transform correlator with significant output images,” Opt. Eng. 40, 1584–1589 (2001) [CrossRef]

, 8

8. H.T. Chang and C.T. Chen, “Enhanced optical image verification based on joint transform correlator adopting Fourier hologram,” Opt. Rev. 11, 165–169 (2004) [CrossRef]

]. In order to reconstruct asymmetric images at the output plane, a complex or, at least, a phase function is required such that the magnitude of its Fourier transform can be asymmetric. Chang and Chen [9

9. H.T. Chang and Ching T. Chen, “Asymmetric-image verification for security optical systems based on joint transform correlator architecture,” Opt. Commun. 239, 43–54 (2004) [CrossRef]

] proposed an optical asymmetric-image verification system based on the joint transform correlator architecture. A complex function at the frequency domain can be obtained by attaching an additional phase mask H 3(u,v) = exp[i2πP 3(u,v)] to the SLM that displays the joint power spectrum, which is shown in the amplitude form. The first phase key h 1(x,y) at the input plane and this attached phase mask, as the second phase key H 3(u, v), are pairwised and iteratively retrieved by the use of the POCS and the multiple phase retrieval algorithms [10

10. H.T. Chang, W. C. Lu, and C. J. Kuo, “Multiple-phase retrieval for optical security systems using random phase encoding,” Appl. Opt. 41, 4825–4834 (2002) [CrossRef] [PubMed]

]. The reconstruction of asymmetric images is thus possible and the system security is also improved because two phase keys are required in reconstructing the target image.

Fig. 2. The schematic diagram of the proposed architecture.

However, another drawback is that the second phase key H 3(u, v) attached at the SLM dominates the reconstructed image quality [9

9. H.T. Chang and Ching T. Chen, “Asymmetric-image verification for security optical systems based on joint transform correlator architecture,” Opt. Commun. 239, 43–54 (2004) [CrossRef]

] because (1) the phase information is usually more important than the amplitude information in signal reconstruction and (2) the data amount of the phase function H 3(u, v) is much larger than that of the phase function h 1(x, y) at the input plane. Therefore, even two phase keys belong to different iterations, the resulting images that are obtained from using the mismatched phase keys still show significant meaning with the target image, especially for binary images. That is, any of the retrieved phase attached at the SLM dominates the quality of the reconstructed image. It could be a disadvantage of this method because the recovered image may be visible when mismatched keys are used. On the other hand, to current techniques, SLMs usually can only display either the amplitude or phase information with a limited resolution. Moreover, additional phase mask increases the extra cost and the alignment requirement in the optical system. Therefore, it is desirable to perform asymmetric-image verification using only a single SLM in the reconstruction stage, which is the case of traditional JTC architecture.

Fig. 3. The block diagram of the proposed method.

2 The Proposed Architecture

Figure 2 shows the schematic diagram of the proposed asymmetric image verification system based on the fully-phase JTC architecture. Compared with the conventional JTC architecture for image verification, there are three major differences: (1) The detected real and non-negative joint power spectrum SP(u, v) is converted to the phase information exp[iSP(u, v)] for further processing. (2) A nonlinear transformation is applied to the converted phase information such that a more broaden histogram of the transformed phase information exp[iSI(u, v)] can be obtained. On the other hand, the converted phase information will vary a lot when the input phase key h 1(x, y) is changed. (3) An extra phase key exp[iH 3(u, v)], which is required in the proposed method and can further increase the system security level, is employed. The first two operations shown above can be performed by the use of digital signal processing techniques in a computer. On the other hand, the extra phase key exp[iH 3(u, v)] is determined by the use of the retrieved phase exp[iO(u, v)] and the transformed phase exp[iSI(u, v)]. The detailed steps for retrieving the second phase key exp[iH 3(u, v)] based on a given target image o(x, y) and the first phase key h 1(x, y) will be provided.

Figure 3 shows the block diagram of the proposed method. The two phase functions h 1(x, y) and h 2(x, y) at the input plane are fourier transformed and then detected by a square-law detector such as a CCD to obtain the joint power spectrum exp[iSP(u, v)]. That is,

SP(u,v)=FT{h1(x,y)+h2(x,y)}2.
(1)

The joint power spectrum will be further processed in the computer. Given a target image o(x, y) at the output plane, the phase required to be appeared at the SLM can be retrieved by applying the POCS algorithm to the optical 2-f architecture [7

7. J. Rosen, “Learning in correlators based on projection onto constraint sets,” Opt. Lett. 18, 1183–1185 (1993) [CrossRef] [PubMed]

]. Thus the Fourier phase that can reconstruct the target image in the spatial domain can be retrieved. Suppose that the reconstructed image converges to the target image after k iterations. The retrieved phase function is denoted as Ok(u,v) and the corresponding reconstruction image becomes |FT{Ok(u,v)}| = |ok(x,y)| ≈ o(x,y). The quality of reconstructed image can be shown much better than that of the conventional image verification system based on the JTC architecture.

The nonlinear transformation is a critical element in the proposed method. It converts the joint power spectrum into phase information and makes the histogram of the converted phase information more uniformly distributed over the whole range [0,2π], which is similar to that of the phase retrieved from the 2-f optical architecture using POCS algorithm. Two representative nonlinear functions are employed in the proposed method: the power-law and log-sigmoid functions. For a given input signal ξ, the output η of the nonlinear function g is denoted as η = g(ξ). The power-law function is defined as

η=g(ξ)=ξb,
(2)

while the log-sigmoid function is defined as

η=g(ξ)=11+eaξ,
(3)

where both the parameters a and b are constants. Figures 4(a) and 4(b) show the power-law and log-sigmoid functions under different parameters, respectively. Note that the ranges of the power spectrum and phase information should be normalized, respectively, to the input and output of the nonlinear functions.

Fig. 4. Two representative nonlinear functions: (a) power law: η = ξb; (b) Log-sigmoid η=11+eaξ.

Here the second phase key exp[iH 3(u, v)] is thus required to make the retrieved phase information Ok(u,v) appear at the SLM. Thus the target image can be reconstructed at the output plane. The second phase key exp[iH 3(u, v)] is determined as

eiH3(u,v)=eiOk(u,v)eiSI(u,v)=ei[Ok(u,v)SI(u,v)],
(4)

where

H3(u,v)={Ok(u,v)SI(u,v),ifOk(u,v)SI(u,v)Ok(u,v)SI(u,v)+2π,ifOk(u,v)<SI(u,v).
(5)

Fig. 5. (a) The symmetric and (b) asymmetric test imgaes.

3 Simulation Results

In computer simulation, the quality of the reconstructed images of size B × B pixels are represented by the mean squared error (MSE), which is defined as

MSE[o(x,y),ok(x,y)]=1B2x=0B1y=0B1[o(x,y)ok(x,y)]2.
(6)

Figures 5(a) and 5(b) show the symmetric and asymmetric target images, respectively, used for testing the proposed method. Figure 6(a) shows the grayscale image of two randomly-generated phase key h 1(x, y) and the phase lock h 2(x, y) at the input plane. Note that the phase information has been transformed to the grayscale values for visualization. Figures 6(b) and 6(c) show the visualized phase information exp[iSP(u, v)] after using the power-law and log-sigmoid functions, respectively. The histograms of Figs. 6(b)–6(d) are shown in Fig. 7(a)–7(c), respectively. Obviously, the histograms of the phase information shown in Figs. 6(a) and 6(b) are more nonuniform and not enough uniformly distributed, respectively. Figure 6(c) shows the much better equalization effect is obtained by the use of the log-sigmoid function with the parameter a = 9. On the other side, the iteration algorithm is used to retrieve the phase information exp[iOk(u, v)] at the SLM for a given target image. Then the second phase key exp[iH(u, v)] can be computed according to Eqs. (4) and (5). Note that the reconstructed image quality depends only on this stage and will not be affected by any randomly generated first phase key h 1(x, y) and the phase lock h 2(x, y).

Fig. 6. (a) Two phase keys at the input plane; (b) The joint power spectrum at the Fourier plane. (c) The power spectrum transformed by the power-law function with the parameter b = 0.3, (d) The power spectrum transformed by the log-sigmoid function with the parameter a = 9.
Fig. 7. (a) The histogram of original joint power spectrum; (b) The histogram of the spectrum after using the power-law transformation; (c) The histogram of the spectrum after using the transform of log-sigmoid function.

Table 1. MSE results under different values of the parameter b in the power-law function in reconstructing symmetric and asymmetric images with correct and wrong phase keys h 1(x, y) and H 3(u, v).

table-icon
View This Table
| View All Tables
Fig. 8. Reconstructed results of the verification system based on the previous optical architecture: (a) symmetric image with the correct phase key h 1(x, y); (b) asymmetric image with the correct phase key h 1(x, y); (c) symmetric image with a wrong phase key h 1(x, y); (d) asymmetric image with a wrong phase key h 1(x, y).

Table 2. MSE results under different values of the parameter a in the log-sigmoid function in reconstructing the image with correct phase keys h 1(x, y) and H 3(u, v).

table-icon
View This Table
| View All Tables

The MSE results of reconstructed symmetric and asymmetric images in the proposed method when two phase keys, h 1(x, y) and H 3(u.v), are correct and when one of the phase keys, h 1(x, y), is wrong are provided for comparison. Table 1 summarizes the MSE results of the proposed method using the power-law function with different values of the parameter b. As shown in this table, the parameter value b = 0.3 can lead almost the best performance for both the symmetric and asymmetric images. On the other hand, Table 2 summarizes the MSE results of the proposed method using the log-sigmoid function with different values of the parameter a. As shown in this table, the parameter value a = 9 can lead almost the best performance for both the symmetric and asymmetric images. From the MSE results shown in both tables, the reconstructed image quality is about the same when both keys are correct. On the other hand, the log-sigmoid function can make the MSE of the reconstructed image much larger than that of using power-law function when the first phase key h 1(x, y) at the input plane is wrong. Therefore, the log-sigmoid function is preferred to as the nonlinear function for transforming the joint power spectrum in the proposed method. Since there are many nonlinear transform functions, different transforms and parameters will lead to different performances on the security level and reconstructed image quality. Furthermore, it will be worthy to explore or investigate other nonlinear transforms for obtaining better performance.

Fig. 9. Reconstructed results of the proposed verification system using the power-law function for converting the joint power spectrum to phase information: (a) symmetric image with the correct phase key h 1(x, y); (b) asymmetric image with the correct phase key h 1(x, y); (c) symmetric image with a wrong phase key h 1(x, y); (d) asymmetric image with a wrong phase key h 1(x, y).
Fig. 10. Reconstructed results of the proposed verification system using the log-sigmoid function for converting the joint power spectrum to phase information: (a) symmetric image with the correct phase keys h 1(x, y) and H 3(u, v); (b) asymmetric image with the correct phase key h 1(x, y) and H 3(u, v); (c) symmetric image with a wrong phase key h 1(x, y) and correct H 3(u, v); (d) asymmetric image with a wrong phase key h 1(x, y) and correct H 3(u, v).

4 Conclusion

References and links

1.

C.J. Weaver and J.W. Goodman, “A technique for optically convolving two functions,” Appl. Opt. 5, 1248–1249 (1966) [CrossRef] [PubMed]

2.

B. Javidi and J.L. Horner, “Optical pattern recognition for validation and security verification,” Opt. Eng. 33, 1752–1756 (1994) [CrossRef]

3.

B. Javidi and T. Nomura, “Polarization encoding for optical security systems,” Opt. Eng. 39, 2439–2443 (2000) [CrossRef]

4.

T. Nomura and B. Javidi, “Optical encryption using a joint transform correlator architecture,” Opt. Eng. 39, 2031–2035 (2000) [CrossRef]

5.

D. Abookasis, O. Arazi, J. Rosen, and B. Javidi, “Security optical systems based on a joint transform correlator with significant output images,” Opt. Eng. 40, 1584–1589 (2001) [CrossRef]

6.

J.R. Fienup, “Phase retrieval algorithm: a comparison,” Appl. Opt. 22, 2758–2769 (1982) [CrossRef]

7.

J. Rosen, “Learning in correlators based on projection onto constraint sets,” Opt. Lett. 18, 1183–1185 (1993) [CrossRef] [PubMed]

8.

H.T. Chang and C.T. Chen, “Enhanced optical image verification based on joint transform correlator adopting Fourier hologram,” Opt. Rev. 11, 165–169 (2004) [CrossRef]

9.

H.T. Chang and Ching T. Chen, “Asymmetric-image verification for security optical systems based on joint transform correlator architecture,” Opt. Commun. 239, 43–54 (2004) [CrossRef]

10.

H.T. Chang, W. C. Lu, and C. J. Kuo, “Multiple-phase retrieval for optical security systems using random phase encoding,” Appl. Opt. 41, 4825–4834 (2002) [CrossRef] [PubMed]

OCIS Codes
(070.2580) Fourier optics and signal processing : Paraxial wave optics
(070.4550) Fourier optics and signal processing : Correlators
(100.3010) Image processing : Image reconstruction techniques
(100.5070) Image processing : Phase retrieval

ToC Category:
Fourier Optics and Optical Signal Processing

History
Original Manuscript: December 20, 2005
Revised Manuscript: February 13, 2006
Manuscript Accepted: February 14, 2006
Published: February 20, 2006

Citation
Hsuan Chang and Chao-C. Chen, "Fully-phase asymmetric-image verification system based on joint transform correlator," Opt. Express 14, 1458-1467 (2006)
http://www.opticsinfobase.org/oe/abstract.cfm?URI=oe-14-4-1458


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. C.J. Weaver and J.W. Goodman, "A technique for optically convolving two functions," Appl. Opt. 5, 1248-1249 (1966). [CrossRef] [PubMed]
  2. B. Javidi and J.L. Horner, "Optical pattern recognition for validation and security verification," Opt. Eng. 33, 1752-1756 (1994). [CrossRef]
  3. B. Javidi and T. Nomura, "Polarization encoding for optical security systems," Opt. Eng. 39, 2439-2443 (2000). [CrossRef]
  4. T. Nomura and B. Javidi, "Optical encryption using a joint transform correlator architecture," Opt. Eng. 39, 2031-2035 (2000). [CrossRef]
  5. D. Abookasis, O. Arazi, J. Rosen, and B. Javidi, "Security optical systems based on a joint transform correlator with significant output images," Opt. Eng. 40, 1584-1589 (2001). [CrossRef]
  6. J.R. Fienup, "Phase retrieval algorithm: a comparison," Appl. Opt. 22, 2758-2769 (1982). [CrossRef]
  7. J. Rosen, "Learning in correlators based on projection onto constraint sets," Opt. Lett. 18, 1183-1185 (1993). [CrossRef] [PubMed]
  8. H.T. Chang and C.T. Chen, "Enhanced optical image verification based on joint transform correlator adopting Fourier hologram," Opt. Rev. 11, 165-169 (2004). [CrossRef]
  9. H.T. Chang and ChingT. Chen, "Asymmetric-image verification for security optical systems based on joint transform correlator architecture," Opt. Commun. 239, 43-54 (2004). [CrossRef]
  10. H.T. Chang, W. C. Lu, and C. J. Kuo, "Multiple-phase retrieval for optical security systems using random phase encoding," Appl. Opt. 41, 4825-4834 (2002). [CrossRef] [PubMed]

Cited By

Alert me when this paper is cited

OSA is able to provide readers links to articles that cite this paper by participating in CrossRef's Cited-By Linking service. CrossRef includes content from more than 3000 publishers and societies. In addition to listing OSA journal articles that cite this paper, citing articles from other participating publishers will also be listed.


« Previous Article  |  Next Article »

OSA is a member of CrossRef.

CrossCheck Deposited