Abstract

Lossy image compression techniques allow arbitrarily high compression rates but at the price of poor image quality. We applied maximum likelihood difference scaling to evaluate image quality of nine images, each compressed via vector quantization to ten different levels, within two different color spaces, RGB and CIE 1976 L*a*b*. In L*a*b* space, images could be compressed on average by 32% more than in RGB space, with little additional loss in quality. Further compression led to marked perceptual changes. Our approach permits a rapid, direct measurement of the consequences of image compression for human observers.

© 2007 Optical Society of America

Full Article  |  PDF Article

References

  • View by:
  • |
  • |

  1. A. Gersho and R. M. Gray, Vector Quantization and Signal Compression (Kluwer Academic, 1991).
  2. I. E. G. Richardson, H.264 and MPEG-4 Video Compression: Video Coding for Next-generation Multimedia (Wiley, UK, 2003).
    [CrossRef]
  3. M. Flierl and B. Girod, Video Coding with Superimposed Motion-Compensated Signals: Applications to H.264 and Beyond (Kluwer Academic, 2004).
  4. R. M. Gray, "Vector quantization," IEEE ASSP Mag. 1, 4-29 (1984).
    [CrossRef]
  5. L. T. Maloney and J. N. Yang, "Maximum likelihood difference scaling," J. Vision 3, 573-585 (2003). http://www.journalofvision.org/3/8/5.
    [CrossRef]
  6. G. Obein, K. Knoblauch, and F. Viénot, "Difference scaling of gloss: non-linearity, binocularity, and constancy," J. Vision 4, 711-720 (2004). http://journalofvision.org/4/9/4.
    [CrossRef]
  7. G. Rhodes, L. T. Maloney, J. Turner, and L. Ewing, "Adaptive face coding and discrimination around the average face." Vision Res. 47, 974-989 (2007).
    [CrossRef] [PubMed]
  8. G. Wyszecki and W. S. Stiles, Color Science: Concepts and Methods, Quantitative Data and Formulas, 2nd ed. (Wiley, 1982).
  9. N. Mulroney and M. D. Fairchild, "Color space selection for JPEG image compression," in Proceedings of 1st IS&T/SID Color Imaging Conference (Society for Imaging Science and Technology, 1993), pp. 157-159.
  10. D. M. Green and J. A. Swets, Signal Detection Theory and Psychophysics (Wiley, New York, 1974).
  11. C. Charrier, K. Knoblauch, and H. Cherifi, "Perceptual distortion analysis of color image based coding," Proc. SPIE 3005, 134-143 (1997).
    [CrossRef]
  12. R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria (2007). ISBN 3-900051-07-0. http://www.R-project.org.
  13. P. McCullagh and J. A. Nelder, Generalized Linear Models (Chapman & Hall, 1989).
  14. B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap (Chapman & Hall, New York, 1993).
  15. A. B. Watson and L. Kreslake, "Measurement of visual impairment scales for digital video," Proc. SPIE 4299, 79-89 (2001).
    [CrossRef]
  16. G. Buchsbaum and A. Gottschalk, "Trichromacy, opponent colours coding and optimum colour information transmission in the retina," Proc. R. Soc. London, Ser. B 220, 89-113 (1983).
    [CrossRef]
  17. D. H. Krantz, R. D. Luce, P. Suppes, and A. Tversky, Foundations of Measurement (Academic, 1971), pp. 140-141.
  18. D. H. Krantz, "A theory of context effects based on cross-context matching," J. Math. Psychol. 5, 1-48 (1968).
    [CrossRef]
  19. D. H. Krantz, "A theory of magnitude estimation and cross-modality matching," J. Math. Psychol. 9, 168-199 (1972).
    [CrossRef]
  20. R. N. Shepard, "On the status of 'direct' psychophysical measurement," in Minnesota Studies in the Philosophy of Science, Vol. 9, C.W.Savage, ed. (University of Minnesota Press, 1978), pp. 441-490.
  21. R. N. Shepard, "Psychological relations and psychophysical scales: on the status of 'direct' psychophysical measurement," J. Math. Psychol. 24, 21-57 (1981).
    [CrossRef]
  22. M. C. Boschman, "DifScal: a tool for analyzing difference ratings on an ordinal category scale," Behav. Res. Methods Instrum. Comput. 33, 10-20 (2001).
    [CrossRef] [PubMed]
  23. A. Mood, F. A. Graybill, and D. C. Boes, Introduction to the Theory of Statistics, 3rd ed. (McGraw-Hill, 1974).
  24. H. Wickham, ggplot: An Implementation of the Grammar of Graphics in R (2006). R package version 0.4.0. http://had.co.nz/ggplot.
  25. K. Knoblauch, C. Charrier, H. Cherifi, J. N. Yang, and L. T. Maloney, "Difference scaling of image quality in compression-degraded images," Perception 27, S174 (1998).

2007

G. Rhodes, L. T. Maloney, J. Turner, and L. Ewing, "Adaptive face coding and discrimination around the average face." Vision Res. 47, 974-989 (2007).
[CrossRef] [PubMed]

2004

G. Obein, K. Knoblauch, and F. Viénot, "Difference scaling of gloss: non-linearity, binocularity, and constancy," J. Vision 4, 711-720 (2004). http://journalofvision.org/4/9/4.
[CrossRef]

2003

L. T. Maloney and J. N. Yang, "Maximum likelihood difference scaling," J. Vision 3, 573-585 (2003). http://www.journalofvision.org/3/8/5.
[CrossRef]

2001

A. B. Watson and L. Kreslake, "Measurement of visual impairment scales for digital video," Proc. SPIE 4299, 79-89 (2001).
[CrossRef]

M. C. Boschman, "DifScal: a tool for analyzing difference ratings on an ordinal category scale," Behav. Res. Methods Instrum. Comput. 33, 10-20 (2001).
[CrossRef] [PubMed]

1998

K. Knoblauch, C. Charrier, H. Cherifi, J. N. Yang, and L. T. Maloney, "Difference scaling of image quality in compression-degraded images," Perception 27, S174 (1998).

1997

C. Charrier, K. Knoblauch, and H. Cherifi, "Perceptual distortion analysis of color image based coding," Proc. SPIE 3005, 134-143 (1997).
[CrossRef]

1984

R. M. Gray, "Vector quantization," IEEE ASSP Mag. 1, 4-29 (1984).
[CrossRef]

1983

G. Buchsbaum and A. Gottschalk, "Trichromacy, opponent colours coding and optimum colour information transmission in the retina," Proc. R. Soc. London, Ser. B 220, 89-113 (1983).
[CrossRef]

1981

R. N. Shepard, "Psychological relations and psychophysical scales: on the status of 'direct' psychophysical measurement," J. Math. Psychol. 24, 21-57 (1981).
[CrossRef]

1972

D. H. Krantz, "A theory of magnitude estimation and cross-modality matching," J. Math. Psychol. 9, 168-199 (1972).
[CrossRef]

1968

D. H. Krantz, "A theory of context effects based on cross-context matching," J. Math. Psychol. 5, 1-48 (1968).
[CrossRef]

Behav. Res. Methods Instrum. Comput.

M. C. Boschman, "DifScal: a tool for analyzing difference ratings on an ordinal category scale," Behav. Res. Methods Instrum. Comput. 33, 10-20 (2001).
[CrossRef] [PubMed]

IEEE ASSP Mag.

R. M. Gray, "Vector quantization," IEEE ASSP Mag. 1, 4-29 (1984).
[CrossRef]

J. Math. Psychol.

D. H. Krantz, "A theory of context effects based on cross-context matching," J. Math. Psychol. 5, 1-48 (1968).
[CrossRef]

D. H. Krantz, "A theory of magnitude estimation and cross-modality matching," J. Math. Psychol. 9, 168-199 (1972).
[CrossRef]

R. N. Shepard, "Psychological relations and psychophysical scales: on the status of 'direct' psychophysical measurement," J. Math. Psychol. 24, 21-57 (1981).
[CrossRef]

J. Vision

L. T. Maloney and J. N. Yang, "Maximum likelihood difference scaling," J. Vision 3, 573-585 (2003). http://www.journalofvision.org/3/8/5.
[CrossRef]

G. Obein, K. Knoblauch, and F. Viénot, "Difference scaling of gloss: non-linearity, binocularity, and constancy," J. Vision 4, 711-720 (2004). http://journalofvision.org/4/9/4.
[CrossRef]

Perception

K. Knoblauch, C. Charrier, H. Cherifi, J. N. Yang, and L. T. Maloney, "Difference scaling of image quality in compression-degraded images," Perception 27, S174 (1998).

Proc. R. Soc. London, Ser. B

G. Buchsbaum and A. Gottschalk, "Trichromacy, opponent colours coding and optimum colour information transmission in the retina," Proc. R. Soc. London, Ser. B 220, 89-113 (1983).
[CrossRef]

Proc. SPIE

A. B. Watson and L. Kreslake, "Measurement of visual impairment scales for digital video," Proc. SPIE 4299, 79-89 (2001).
[CrossRef]

C. Charrier, K. Knoblauch, and H. Cherifi, "Perceptual distortion analysis of color image based coding," Proc. SPIE 3005, 134-143 (1997).
[CrossRef]

Vision Res.

G. Rhodes, L. T. Maloney, J. Turner, and L. Ewing, "Adaptive face coding and discrimination around the average face." Vision Res. 47, 974-989 (2007).
[CrossRef] [PubMed]

Other

G. Wyszecki and W. S. Stiles, Color Science: Concepts and Methods, Quantitative Data and Formulas, 2nd ed. (Wiley, 1982).

N. Mulroney and M. D. Fairchild, "Color space selection for JPEG image compression," in Proceedings of 1st IS&T/SID Color Imaging Conference (Society for Imaging Science and Technology, 1993), pp. 157-159.

D. M. Green and J. A. Swets, Signal Detection Theory and Psychophysics (Wiley, New York, 1974).

R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria (2007). ISBN 3-900051-07-0. http://www.R-project.org.

P. McCullagh and J. A. Nelder, Generalized Linear Models (Chapman & Hall, 1989).

B. Efron and R. J. Tibshirani, An Introduction to the Bootstrap (Chapman & Hall, New York, 1993).

R. N. Shepard, "On the status of 'direct' psychophysical measurement," in Minnesota Studies in the Philosophy of Science, Vol. 9, C.W.Savage, ed. (University of Minnesota Press, 1978), pp. 441-490.

D. H. Krantz, R. D. Luce, P. Suppes, and A. Tversky, Foundations of Measurement (Academic, 1971), pp. 140-141.

A. Gersho and R. M. Gray, Vector Quantization and Signal Compression (Kluwer Academic, 1991).

I. E. G. Richardson, H.264 and MPEG-4 Video Compression: Video Coding for Next-generation Multimedia (Wiley, UK, 2003).
[CrossRef]

M. Flierl and B. Girod, Video Coding with Superimposed Motion-Compensated Signals: Applications to H.264 and Beyond (Kluwer Academic, 2004).

A. Mood, F. A. Graybill, and D. C. Boes, Introduction to the Theory of Statistics, 3rd ed. (McGraw-Hill, 1974).

H. Wickham, ggplot: An Implementation of the Grammar of Graphics in R (2006). R package version 0.4.0. http://had.co.nz/ggplot.

Cited By

OSA participates in CrossRef's Cited-By Linking service. Citing articles from OSA journals and other participating publishers are listed here.

Alert me when this article is cited.


Figures (6)

Fig. 1
Fig. 1

Effects of VQ compression. The original image (0% compression) is shown after VQ compression using a codebook based on the LBG algorithm applied to the image encoded in L * a * b * color space (see text). Larger compressions lead to evident decreases in image quality.

Fig. 2
Fig. 2

An example of a single difference scaling trial. On each trial, the observer sees two pairs of images and judges which pair (upper or lower) has the greater perceived difference. The upper pair corresponds to images compressed by factors of 6 (left) and 15 (right), the lower pair to images compressed by factors of 18 (left) and 27 (right).

Fig. 3
Fig. 3

Example of a difference scale. On the horizontal scale we plot degree of image compression γ, and on the vertical we plot difference scale values derived from a psychophysical procedure, MLDS [5]. The difference scale values are estimates based on the observer’s judgments of superthreshold perceptual differences between the images portrayed in Fig. 1. See text. Compression rates up to a factor of 12–15 result in little perceived difference. Above a factor of 15, the difference scale values increase markedly with increased compression rate.

Fig. 4
Fig. 4

Images. The nine images used in the experiments are shown, with mnemonic labels. For each image and each color space, RGB and L * a * b * , we estimated a difference scale based on each observer’s judgments.

Fig. 5
Fig. 5

Difference scales for each image and observer. The difference scales for each image and observer are shown. The image labels correspond to those of Fig. 4. Observers’ initials are indicated in the right-hand labels for each row of panels. In each plot we show the scale corresponding to the VQ compression based on RGB color space [dashed, red (online)] and separately the scale values based on L * a * b * color space (solid, black) The confidence intervals shown ( ± 1 SD) were estimated by a Bootstrap procedure [14] as described by Maloney and Yang [5]. The fitted lines are J-functions, defined in the text.

Fig. 6
Fig. 6

Summary results. (a) Difference scales in Fig. 5 averaged across all images and both color spaces, separately for each of the two observers. Average observer results are in good agreement. (b) Difference scales in Fig. 5 averaged across all images and each observer separately for each color space.

Tables (2)

Tables Icon

Table 1 Estimated Parameters for J-Functions Fitted to Perceptual Scales for Images in L * a * b * Color Space

Tables Icon

Table 2 Estimated Parameters for J-Functions Fitted to Perceptual Scales for Images in RGB Color Space

Equations (17)

Equations on this page are rendered with MathJax. Learn more.

ψ b ψ a > ψ d ψ c ;
δ ( a , b ; c , d ) = L c d L a b = ψ d ψ c ψ b ψ a .
Δ ( a , b ; c , d ) = δ ( a , b ; c , d ) + ϵ = L c d L a b + ϵ ,
Δ ( a , b ; c , d ) > 0 .
L ( Ψ , σ ) = k = 1 P Φ ( δ ( q k ) σ ) 1 R k ( 1 Φ ( δ ( q k ) σ ) ) R k ,
ψ ( γ ) = a 1 γ + a 3 , γ B ,
= a 2 ( γ B ) + a 1 B + a 3 , γ > B .
δ = ψ d ψ c ψ b ψ a ,
Δ = δ + ϵ ,
δ = ψ d ψ c ψ b + ψ a ,
Δ = δ + ϵ ,
( 1 3 ; 5 7 7 9 ; 4 5 1 6 ; 7 8 3 4 ; 9 10 ) ( 0 1 0 1 0 1 0 0 0 0 0 1 1 0 1 0 1 0 0 0 0 0 1 1 1 0 0 0 1 1 0 0 0 0 1 1 ) .
δ = M Ψ ̃ ,
p = Φ ( δ )
P [ R = 1 ] = Φ ( M Ψ ̃ ) ,
E [ R ] = Φ ( M Ψ ̃ ) ,
Φ 1 ( E [ R ] ) = δ = M Ψ ̃ .

Metrics