Skip to main content

Low-complexity rounded KLT approximation for image compression


The Karhunen–Loève transform (KLT) is often used for data decorrelation and dimensionality reduction. Because its computation depends on the matrix of covariances of the input signal, the use of the KLT in real-time applications is severely constrained by the difficulty in developing fast algorithms to implement it. In this context, this paper proposes a new class of low-complexity transforms that are obtained through the application of the round function to the elements of the KLT matrix. The proposed transforms are evaluated considering figures of merit that measure the coding power and distance of the proposed approximations to the exact KLT and are also explored in image compression experiments. Fast algorithms are introduced for the proposed approximate transforms. It was shown that the proposed transforms perform well in image compression and require a low implementation cost.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7

Availability of data and material

Not applicable.

Code availability

Not applicable.


  1. 1.

    Ahmed, N., Natarajan, T., Rao, K.R.: Discrete cosine transfom. IEEE Trans. Comput. 100, 90–93 (1974).

    Article  MATH  Google Scholar 

  2. 2.

    Bayer, F.M., Cintra, R.J.: DCT-like transform for image compression requires 14 additions only. Electron. Lett. 48(15), 919–921 (2012).

    Article  Google Scholar 

  3. 3.

    Bayer, F.M., Cintra, R.J., Edirisuriya, A., Madanayake, A.: A digital hardware fast algorithm and FPGA-based prototype for a novel 16-point approximate DCT for image compression applications. Meas. Sci. Technol. 23(11), 114010 (2012)

    Article  Google Scholar 

  4. 4.

    Betzel, F., Khatamifard, K., Suresh, H., Lilja, D.J., Sartori, J., Karpuzcu, U.: Approximate communication: Techniques for reducing communication bottlenecks in large-scale parallel systems. ACM Comput. Surv. (CSUR) 51(1), 1–32 (2018).

    Article  Google Scholar 

  5. 5.

    Blahut, R.E.: Fast Algorithms for Signal Processing. Cambridge University Press, Cambridge (2010)

    Book  Google Scholar 

  6. 6.

    Bouguezel, S., Ahmad, M.O., Swamy, M.: Low-complexity \(8 \times 8\) transform for image compression. Electron. Lett. 44(21), 1249–1250 (2008).

    Article  Google Scholar 

  7. 7.

    Bouguezel, S., Ahmad, M.O., Swamy, M.N.S.: A low-complexity parametric transform for image compression. In: IEEE International Symposium of Circuits and Systems (ISCAS), pp. 2145–2148 (2011).

  8. 8.

    Britanak, V., Yip, P.C., Rao, K.R.: Discrete Cosine and Sine Transforms: General Properties, Fast Algorithms and Integer Approximations. Academic Press, New York (2010)

    Google Scholar 

  9. 9.

    Canterle, D.R., da Silveira, T.L., Bayer, F.M., Cintra, R.J.: A multiparametric class of low-complexity transforms for image and video coding. Signal Process. 176, 107685 (2020).

    Article  Google Scholar 

  10. 10.

    Chen, J., Liu, S., Deng, G., Rahardja, S.: Hardware efficient integer discrete cosine transform for efficient image/video compression. IEEE Access 7, 152635–152645 (2019).

    Article  Google Scholar 

  11. 11.

    Cintra, R.J.: An integer approximation method for discrete sinusoidal transforms. Circuits Syst. Signal Process. 30(6), 1481 (2011).

    MathSciNet  Article  MATH  Google Scholar 

  12. 12.

    Cintra, R.J., Bayer, F.M.: A DCT approximation for image compression. IEEE Signal Process. Lett. 18(10), 579–582 (2011).

    Article  Google Scholar 

  13. 13.

    Cintra, R.J., Bayer, F.M., Tablada, C.J.: Low-complexity 8-point DCT approximations based on integer functions. Signal Process. 99, 201–214 (2014).

    Article  Google Scholar 

  14. 14.

    Coelho, D.F., Cintra, R.J., Madanayake, A., Perera, S.M.: Low-complexity scaling methods for DCT-II approximations. IEEE Trans. Signal Process. (2021).

    MathSciNet  Article  Google Scholar 

  15. 15.

    Fan, K., Wang, R., Lin, W., Duan, L., Gao, W.: Signal-independent separable KLT by offline training for video coding. IEEE Access 7, 33087–33093 (2019).

    Article  Google Scholar 

  16. 16.

    Goldston, D.: Big data: data wrangling. Nat. News 455(7209), 15 (2008).

    Article  Google Scholar 

  17. 17.

    Gonzalez, R.C., Woods, R.E.: Digital Image Processing, 3rd edn. Prentice Hall, Upper Saddle River (2006)

    Google Scholar 

  18. 18.

    Harville, D.A.: Trace of a (square) matrix. In: Matrix Algebra from a Statistician’s Perspective, pp. 49–53. Springer (1997)

    Chapter  Google Scholar 

  19. 19.

    Haweel, T.I.: A new square wave transform based on the DCT. Signal Process. 81(11), 2309–2319 (2001).

    Article  MATH  Google Scholar 

  20. 20.

    Higham, N.J.: Functions of Matrices: Theory and Computation, vol. 104. SIAM, Philadelphia (2008)

    Book  Google Scholar 

  21. 21.

    Huang, J., Kumar, T.N., Almurib, H.A., Lombardi, F.: A deterministic low-complexity approximate (multiplier-less) technique for DCT computation. IEEE Trans. Circuits Syst. I Regul. Pap. 66(8), 3001–3014 (2019).

    MathSciNet  Article  MATH  Google Scholar 

  22. 22.

    Huynh-Thu, Q., Ghanbari, M.: Scope of validity of PSNR in image/video quality assessment. Electron. Lett. 44(13), 800–801 (2008).

    Article  Google Scholar 

  23. 23.

    Jain, A.K.: A fast Karhunen–Loève transform for a class of random processes. IEEE Trans. Commun. 24(9), 1023–1029 (1976).

    Article  MATH  Google Scholar 

  24. 24.

    Jolliffe, I.T.: Principal Component Analysis. Springer, New York (1986)

    Book  Google Scholar 

  25. 25.

    Jridi, M., Alfalou, A., Meher, P.K.: A generalized algorithm and reconfigurable architecture for efficient and scalable orthogonal approximation of DCT. IEEE Trans. Circuits Syst. I Regul. Pap. 62(2), 449–457 (2014)

    Article  Google Scholar 

  26. 26.

    Jridi, M., Alfalou, A., Meher, P.K.: A generalized algorithm and reconfigurable architecture for efficient and scalable orthogonal approximation of DCT. IEEE Trans. Circuits Syst. I Regul. Pap. 62(2), 449–457 (2015)

    Article  Google Scholar 

  27. 27.

    Karhunen, K.: Under lineare methoden in der wahr scheinlichkeitsrechnung. Annales Academiae Scientiarun Fennicae Series A1 Mathematia Physica 47 (1947)

  28. 28.

    Katto, J., Komatsu, K., Yasuda, Y.: Short-tap and linear-phase PR filter banks for subband coding of images. In: Visual Communications and Image Processing’92, vol. 1818, pp. 735–746. International Society for Optics and Photonics (1992).

  29. 29.

    Lengwehasatit, K., Ortega, A.: Scalable variable complexity approximate forward DCT. IEEE Trans. Circuits Syst. Video Technol. 14(11), 1236–1248 (2004).

    Article  Google Scholar 

  30. 30.

    Loève, M.: Fonctions aléatoires de second ordre. Processus Stochastique et Mouvement Brownien, pp. 366–420 (1948)

  31. 31.

    Ochoa-Dominguez, H., Rao, K.R.: Discrete Cosine Transform. CRC Press, Boca Raton (2019)

    Book  Google Scholar 

  32. 32.

    Oliveira, R.S., Cintra, R.J., Bayer, F.M., da Silveira, T.L., Madanayake, A., Leite, A.: Low-complexity 8-point DCT approximation based on angle similarity for image and video coding. Multidimens. Syst. Signal Process. 30(3), 1363–1394 (2019).

    MathSciNet  Article  MATH  Google Scholar 

  33. 33.

    Pennebaker, W.B., Mitchell, J.L.: JPEG: Still Image Data Compression Standard. Van Nostrand Reinhold, New York (1992)

    Google Scholar 

  34. 34.

    Potluri, U.S., Madanayake, A., Cintra, R.J., Bayer, F.M., Kulasekera, S., Edirisuriya, A.: Improved 8-point approximate DCT for image and video compression requiring only 14 additions. IEEE Trans. Circuits Syst. I Regul. Pap. 61(6), 1727–1740 (2014).

    Article  Google Scholar 

  35. 35.

    Pourazad, M.T., Doutre, C., Azimi, M., Nasiopoulos, P.: HEVC: The new gold standard for video compression: how does HEVC compare with H.264/AVC? IEEE Consum. Electron. Mag. 1(3), 36–46 (2012).

    Article  Google Scholar 

  36. 36.

    Puchala, D.: Approximate calculation of \(8\)-point DCT for various scenarios of practical applications. EURASIP J. Image Video Process. 2021(1), 1–34 (2021).

  37. 37.

    Puri, A., Chen, X., Luthra, A.: Video coding using the H.264/MPEG-4 AVC compression standard. Signal Process. Image Commun. 19(9), 793–849 (2004).

    Article  Google Scholar 

  38. 38.

    Ray, W., Driver, R.: Further decomposition of the Karhunen–Loève series representation of a stationary random process. IEEE Trans. Inf. Theory 16(6), 663–668 (1970).

    Article  MATH  Google Scholar 

  39. 39.

    Reed, I.S., Lan, L.: A fast approximate Karhunen–Loève transform (AKLT) for data compression. J. Vis. Commun. Image Represent. 5(4), 304–316 (1994).

    Article  Google Scholar 

  40. 40.

    Salomon, D.: Data Compression: The Complete Reference. Springer Science and Business Media, Berlin (2004)

    MATH  Google Scholar 

  41. 41.

    Sayood, K.: Introduction to Data Compression. Morgan Kaufmann (2017)

    MATH  Google Scholar 

  42. 42.

    Seber, G.A.: A Matrix Handbook for Statisticians, vol. 15. Wiley, New York (2008)

    MATH  Google Scholar 

  43. 43.

    da Silveira, T.L., Oliveira, R.S., Bayer, F.M., Cintra, R.J., Madanayake, A.: Multiplierless 16-point DCT approximation for low-complexity image and video coding. SIViP 11(2), 227–233 (2017).

    Article  Google Scholar 

  44. 44.

    Singhadia, A., Bante, P., Chakrabarti, I.: A novel algorithmic approach for efficient realization of 2-D-DCT architecture for HEVC. IEEE Trans. Consum. Electron. 65(3), 264–273 (2019).

    Article  Google Scholar 

  45. 45.

    Suzuki, T., Ikehara, M.: Integer DCT based on direct-lifting of DCT-IDCT for lossless-to-lossy image coding. IEEE Trans. Image Process. 19(11), 2958–2965 (2010).

    MathSciNet  Article  MATH  Google Scholar 

  46. 46.

    Takala, J., Nikara, J.: Unified pipeline architecture for discrete sine and cosine transforms of type IV. In: Proceedings of the 3rd International Conference on Information Communication and Signal Processing, vol. 10 (2001)

  47. 47.

    The USC-SIPI image database. Signal and Image Processing Institute of the University of Southern California. (2017)

  48. 48.

    Wallace, G.K.: The JPEG still picture compression standard. IEEE Trans. Consum. Electron. 38(1), xviii–xxxiv (1992).

  49. 49.

    Wang, Z., Bovik, A.C.: Mean squared error: love it or leave it? a new look at signal fidelity measures. IEEE Signal Process. Mag. 26(1), 98–117 (2009).

    Article  Google Scholar 

  50. 50.

    Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE Trans. Image Process. 13(4), 600–612 (2004).

    Article  Google Scholar 

  51. 51.

    Zidani, N., Kouadria, N., Doghmane, N., Harize, S.: Low complexity pruned DCT approximation for image compression in wireless multimedia sensor networks. In: 2019 5th International Conference on Frontiers of Signal Processing (ICFSP), pp. 26–30. IEEE (2019).

Download references


We gratefully acknowledge partial financial support from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) and Fundação de Amparo à Ciência e Tecnologia de Pernambuco (FACEPE), Brazil.


Not applicable.

Author information



Corresponding author

Correspondence to Anabeth P. Radünz.

Ethics declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix: 2D transformation and quantization step

Appendix: 2D transformation and quantization step

Let \({\mathbf {A}}\) be a \(8 \times 8\) sub-block from an image. The 2D transformation from \({\mathbf {A}}\) induced by an approximation \(\widehat{{\mathbf {T}}}\) is given by:

$$\begin{aligned} {\mathbf {B}}&= {\left\{ \begin{array}{ll} \widehat{{\mathbf {T}}} \cdot {\mathbf {A}} \cdot {\widehat{{\mathbf {T}}}}^\top , &{} \text {if }{\mathbf {T}}\text { is orthogonal,} \\ \widehat{{\mathbf {T}}} \cdot {\mathbf {A}} \cdot {\widehat{{\mathbf {T}}}}^{-1}, &{} \text {if }{\mathbf {T}}\text { is non-orthogonal,} \end{array}\right. } \nonumber \\&= {\left\{ \begin{array}{ll} ({\mathbf {u}} \cdot {\mathbf {u}}^\top ) \odot ({\mathbf {T}} \cdot {\mathbf {A}} \cdot {\mathbf {T}}^\top ), &{} \text {if }{\mathbf {T}}\text { is orthogonal,} \\ ({\mathbf {u}} \cdot {\mathbf {v}}^\top ) \odot ({\mathbf {T}} \cdot {\mathbf {A}} \cdot {\mathbf {T}}^{-1}), &{} \text {if }{\mathbf {T}}\text { is non-orthogonal,} \end{array}\right. } \nonumber \\&= {\mathbf {R}} \odot \widehat{{\mathbf {B}}}, \end{aligned}$$

where \({\mathbf {u}} = {\text {diag}}({\mathbf {S}})\) and \({\mathbf {v}}\) is given by the inverse elements from \({\mathbf {u}}\). In the context of JPEG-like compression [48], the quantized coefficient matrix \(\widehat{{\mathbf {B}}}\) is given by:

$$\begin{aligned} \bar{{\mathbf {B}}} = {\text {round}}({\mathbf {B}} \div {\mathbf {Q}}), \end{aligned}$$

where \({\mathbf {Q}}\) is a quantization matrix and \(\div\) denotes the element-wise matrix division.

By applying Eq. (3) in (4), we obtain

$$\begin{aligned} \bar{{\mathbf {B}}} = {\text {round}}({\mathbf {R}} \odot \widehat{{\mathbf {B}}} \div {\mathbf {Q}}) = {\text {round}}(\widehat{{\mathbf {B}}} \div \tilde{{\mathbf {Q}}}), \end{aligned}$$

where \(\tilde{{\mathbf {Q}}} = {\mathbf {Q}} \div {\mathbf {R}}\). Note that \({\mathbf {R}}\) can be absorbed in the quantization step, thus, the complexity of matrix \({\mathbf {S}}\) can be dismissed in the image compression applications [17, 40, 41].

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Radünz, A.P., Bayer, F.M. & Cintra, R.J. Low-complexity rounded KLT approximation for image compression. J Real-Time Image Proc (2021).

Download citation


  • Approximate KLT
  • Image compression
  • Karhunen–Loève transform
  • Low-complexity transforms