OSA's Digital Library

Applied Optics

Applied Optics

APPLICATIONS-CENTERED RESEARCH IN OPTICS

  • Vol. 15, Iss. 11 — Nov. 1, 1976
  • pp: 2846–2854

Scene power spectra: the moment as an image quality merit factor

Norman B. Nill  »View Author Affiliations


Applied Optics, Vol. 15, Issue 11, pp. 2846-2854 (1976)
http://dx.doi.org/10.1364/AO.15.002846


View Full Text Article

Enhanced HTML    Acrobat PDF (1116 KB)





Browse Journals / Lookup Meetings

Browse by Journal and Year


   


Lookup Conference Papers

Close Browse Journals / Lookup Meetings

Article Tools

Share
Citations

Abstract

Coherent optical power spectra of random scene imagery can become a powerful image evaluation technique when properly analyzed. A normalized low order moment of power spectra is set forth as a good image quality merit factor, which also has the advantageous property of taking out the spectrum analyzer’s aperture effect. In addition to the relevant theory, experimental results with this merit factor of power spectra are given that demonstrate a high degree of correlation with subjective quality rankings, as well as the accurate assessment of optimum focus.

© 1976 Optical Society of America

History
Original Manuscript: May 29, 1976
Published: November 1, 1976

Citation
Norman B. Nill, "Scene power spectra: the moment as an image quality merit factor," Appl. Opt. 15, 2846-2854 (1976)
http://www.opticsinfobase.org/ao/abstract.cfm?URI=ao-15-11-2846


Sort:  Author  |  Year  |  Journal  |  Reset  

References

  1. N. Jensen, Photogramm. Eng. 39, 1321 (1973).
  2. At present there is no standard terminology in this area. The term power spectrum will be used in this paper, although other terms used in the literature are spectral power density, spectral density function, energy spectrum, Wiener-Khinchin spectrum, and Wiener spectrum. A good start at standardization is in the distinctly separate definitions given to power spectrum and Wiener spectrum by Thiry, J. Photogr. Sci. 11, 69, 1963.
  3. G. G. Lendaris, G. L. Stanley, IEEE Proc. 58, 198 (1970).
  4. L. J. Cutrona, E. N. Leith, C. J. Palermo, L. J. Porcello, IRE Trans. Inf. Theory IT-6, 386 (1960).
  5. A. Vander Lugt, IEEE Proc. 62, 1300 (1974).
  6. H. Stark, W. R. Bennett, M. Arm, Appl. Opt. 8, 2165 (1969).
  7. H. Stark, B. DimiTriadis, J. Opt. Soc. Am. 65, 425 (1975).
  8. J. W. Goodman, Introduction to Fourier Optics (McGraw-Hill, New York, 1968).
  9. The film’s light amplitude transmittance is actually a complex function, e.g., t(x,y) = h(x,y) exp[jθ(x,y)], where h(x,y) is the positive square root of the light intensity transmittance, and θ(x,y) describes a phase function caused by the emulsion relief image (signal-dependent), as well as film base thickness and index of refraction variations (signal-independent) [A. L. Ingalls, Photogr. Sci. Eng. 4, 135 (1960); H. M. Smith, J. Opt. Soc. Am. 62, 802 (1972)]. The use of an index of refraction matching solution surrounding the film would minimize the first two phase components mentioned. However, this is an impractical technique to use for film that often must remain of archival quality. The emulsion relief image can be considered as a semisignal source since it nonlinearly tracks h(x,y). Experimental results demonstrate that the other two phase components have a relatively minimal deleterious effect upon the measured power spectra. For these reasons, liquid immersion of the film was not used.
  10. See Ref. 7, Eq. (5).
  11. A. Papoulis, Systems and Transforms with Applications in Optics (McGraw-Hill, New York, 1968), p. 140.
  12. By evaluating the following Fourier transform relation [given in R. Bracewell, The Fourier Transform and its Applications (McGraw-Hill, New York, 1965), p. 245](2πju)m(2πjv)nP(u,v)↔F.T.(∂∂x)m(∂∂y)np(x,y)at x = y = 0, we have the relation for two-dimensional rectangular coordinate joint moments of order m + n:∫-∞∞∫-∞∞umvnP(u,v)dudv=1(2πj)m+n(∂∂x)m (∂∂y)np(0,0).Solving this moment equation for (m = 0, n = 2) and (m = 2, n = 0) and combining results give∫-∞∞∫-∞∞(u2+v2)P(u,v)dudv=-14π2[∂2p(x,y)dx2+∂2p(x,y)∂y2]x=y0,which is often called the moment of inertia and is the two-dimensional analog of Eq. (6).
  13. E. H. Linfoot, J. Opt. Soc. Am. 46, 740 (1957).
  14. W. Smith, Modern Optical Engineering (McGraw-Hill, New York, 1966), p. 311.
  15. Y. Itakura, S. Tsutsumi, T. Takagi, Infrared Phys. 14, 17 (1974).
  16. T. Takagi, S. Tutumi, J. Electr. Commun. Jpn. 51-C, 82 (1968).
  17. D. R. Cox, P. A. W. Lewis, Statistical Analysis of Series of Events (Wiley, New York, 1966).
  18. D. Halford, IEEE Proc. 56, 251 (1968).
  19. See Ref. 16 and also note the following. If the scene spectrum falls off only as ν−2, the theoretical existence of∫0∞ν3P(ν)dνat first appears to be less assured (see Bracewell, Ref. 12, p. 141). However, for this case P(ν) will actually fall off faster than ν−2 because of the effect of the sensor system’s spatial frequency response. In any case, over the finite limits actually used, the moment will exist in practice.
  20. K. Stumpf, R. Warren, “Wiener Spectra Flight Test Program,” Itek Corp. Technical Report, Lexington, Mass. (March1974), Vols. 1, 2.
  21. J. J. DePalma, E. M. Lowry, J. Opt. Soc. Am. 52, 328 (1962).
  22. E. M. Granger, K. N. Cupery, Photogr. Sci. Eng. 16, 221 (1972) [substituting f−1df for d(Log f) in the author’s second equation and using the relation (input power spectrum) |τ(f,θ)|2 = output power spectrum, where input power spectrum = f−m, m a constant].
  23. C. Helstrom, Statistical Theory of Signal Detection (Pergamon, New York, 1968), p. 4.
  24. A. H. Katz, J. Opt. Soc. Am. 38, 604 (1948).
  25. F. G. Back, J. Opt. Soc. Am. 43, 685 (1953).
  26. The experimental result that the power spectra of the various measuring apertures closely agree with the theoretically predicted spectra [via Eq. (1)] confirms the working validity of the power spectrum relationship given in Eq. (1).
  27. It is known that the visual ranking technique described is scale-dependent, i.e., all other factors being equal, a photograph with a low scale will have a higher rating than a photograph with a higher scale. On the other hand it can be shown that for physically reasonable forms of F(ν), the normalized low order moment of F(ν) is essentially scale-independent. In order for M to track properly the visual ranking therefore, M is made to be scale-dependent by the conversionM(cycles/m)2=1×106M(cycles/mm)2(Scale)2
  28. See Ref. 14, p. 320.
  29. J. F. Walkup, R. C. Choens, Opt. Eng. 13, 258 (1974).
  30. A. S. Husain-Abidi, Pattern Recognition (Pergamon, New York, 1973), 5.3.
  31. J. Arsac, Fourier Transforms and the Theory of Distributions (Prentice-Hall, Englewood Cliffs, N.J., 1966), p. 32.
  32. J. Stone, Radiation and Optics (McGraw-Hill, New York, 1963), p. 325.

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