access icon free Spatially adaptive image denoising using inter-scale dependence in directionlet domain

© The Institution of Engineering and Technology
Received 31/03/2013, Accepted 12/07/2014, Revised 06/07/2014, Published 14/08/2014

The performance of image processing algorithms can be significantly improved by the application of multi-resolution image representation with directional features. Directionlet transform (DT) is one such representation which has gained popularity over the past few years as an anisotropic, perfect reconstruction and critically sampled basis function with directional vanishing moments along any two directions. In this study, the authors propose a spatially adaptive image denoising scheme for Gaussian noise based on DT by considering the dependences of the directionlet coefficients across different scales. The image is first decomposed using DT and the coefficients so obtained are modelled using a bivariate heavy tailed ‘pdf’ with a local variance parameter to account for inter- and intra-scale dependencies of the coefficients. The DT is made adaptive to the local dominant directions in the image by identifying the dominant directions in the spatially segmented image through the computation of a parameter called ‘directional variance’. Bayesian ‘maximum a posteriori’ estimator is then used to compute the noise free coefficients from the bivariate models of the signal and noise. The denoised image is obtained from the transform coefficients, which were modified using the bivariate shrinkage function, using directional information and inverse DT. Experimental results show that the bivariate shrinkage in directionlet domain achieves better performance than that in wavelet domain, in terms of numerical and perceptual quality.

Inspec keywords: image representation; transforms; image segmentation; image denoising; image resolution; Gaussian noise

Other keywords: image segmentation; noise free coefficients; directional features; DT; spatially adaptive image denoising; directionlet domain; Gaussian noise; transform coefficients; sampled basis function; directional variance; interscale dependence; multiresolution image representation; image processing algorithms; directional vanishing moments; directionlet coefficients; bivariate models; bivariate shrinkage function

Subjects: Computer vision and image processing techniques; Other topics in statistics; Integral transforms; Other topics in statistics; Optical, image and video signal processing; Integral transforms

References

    1. 1)
    2. 2)
      • 12. Strack, J.L., Candes, E.J., Donoho, D.L.: ‘The curvelet transform for image denoising’, IEEE Trans. Image Process., 2000, 11, (6), pp. 670–684.
    3. 3)
    4. 4)
    5. 5)
    6. 6)
    7. 7)
    8. 8)
    9. 9)
    10. 10)
    11. 11)
    12. 12)
      • 20. Jayachandra, D., Makur, A.: ‘Directional variance: a measure to find the directionality in a given image segment’. Proc. of IEEE Int. Symp. on Circuits and Systems, 2010, pp. 1551–1554.
    13. 13)
    14. 14)
    15. 15)
    16. 16)
      • 2. Simoncelli, E.P.: ‘Bayesian denoising of visual images in the wavelet domain’, in Müller, P., Vidakovic, B. (Eds.): ‘Bayesian inference in wavelet based models’ (Springer-Verlag, New York, 1999).
    17. 17)
    18. 18)
    19. 19)
    20. 20)
http://iet.metastore.ingenta.com/content/journals/10.1049/iet-ipr.2014.0112
Loading

Related content

content/journals/10.1049/iet-ipr.2014.0112
pub_keyword,iet_inspecKeyword,pub_concept
6
6
Loading