SLP Header

Joint Denoising and Demosaicking of Noisy Bayer sampled Color Images with LSLCD and Noise Level Estimation

IJCSEC Front Page

Demosaicking and zooming are important for the quality of digital images in resource-constrained single chip devices such as wireless camera phones and vision-based portable devices. During Demosaicking process color artifacts are introduced which can be magnified during zooming process and vice versa. This paper represents a joint demosaicking– zooming scheme by exploiting the fact that red/green and blue/green color difference signals are much smoother than the red, green and blue original signals. Using the high spectral– spatial correlations in the color filter array (CFA) image the color difference signals are computed. The computed color difference signals are used to recover the green channel. This green channel is enlarged using an edge guided zooming scheme. By adding the correspondingly enlarged red/green and blue/green color difference images to the enlarged green channel the enlarged red and blue channels can be found. The proposed joint demosaicking–zooming scheme performs well in both visual perception and Peak Signal to Noise Ratio (PSNR) measurement, reducing much color artifacts arising from demosaicking as well zippers and rings arising from zooming.
Keywords:Demosaicking, Peak Signal to Noise Ratio (PSNR)
Most digital still/video cameras capture images using a single chip and a color filter array (CFA) known as the Bayer color filter array (CFA) pattern. The process to reproduce the color components and enlarge the images while maintaining image quality is crucial to many resource-constrained digital imaging devices such vision-based pocket devices, camera phones and low-cost color value video cameras.
The enlarged image quality depends on two image processing operations: demosaicking (spectral interpolation) and zooming Spatial interpolation). A single chip camera sample only one of the three primary color components (red, green and blue), a full color image is created through a process called demosaicking, which is used to bring out the other two missing color (components at each pixel location. The input image is been enlarged using zooming and spatial interpolation sequentially. The full color is obtained by, enlarged image from a CFA image, i.e. To increase the spatial and spectral resolution of a CFA image, there are three strategies: the CFA image is first demosaicked and then enlarged; second the CFA image is first enlarged and then demosaicked; the demosaicking and zooming are implemented simultaneously. The first two strategies, the quality of the resulted image depends on the demosaicking and zooming algorithms used and their application order to the CFA image. However, errors in the form of color artifacts are usually introduced into the final image.
The Fig.1 shows the flowchart of the proposed joint demosaicking–zooming scheme. Because of the high spectral correlation between color channels, the R–G and B–G color difference signals are much smoother than the original color signals G, R and B. Based on this observation, the R–G and B– G color difference images are estimated and they heavily affect the demosaicking–zooming results, as enlarging the color difference images instead of R and B images significantly reduces the demosaicking and interpolation errors. Once the R– G and B–G values at the red and blue sample positions are estimated, the fully populate R–G and B– G images are used to demosaick the green channel. Since the sampling frequency of green channel is as twice as that of red or blue channel, it can be demosaicked and interpolated more accurately than red and blue channels. The green channel contains the most of the details of an image and then the green channel should serve as an anchor for recovering red and blue channels.


  1. J.E. Adams, Intersections between color plane interpolation and other image processing functions in electronic photography, Proceedings of SPIE 2416 (1995) 144–151.
  2. J.E. Adams, Design of practical color filter array interpolation algorithms for digital cameras, Proceedings of SPIE 3028 (1997) 117– 125.
  3. S. Carrato, L. Tenze, A high quality 2 · image interpolator, IEEE Signal Processing Letters 7 (2000) 132–135.
  4. J.F. Hamilton Jr., J.E. Adams, Adaptive color plane interpolation in single sensor color electronic camera, US Patent, 5 629 734, 1997.
  5. K. Hirakawa, T.W. Parks, Adaptive homogeneitydirected demosaicing algorithm, IEEE Transactions on Image Processing 14 (2005) 360–369.
  6. R. Kimmel, Demosaicing: Image reconstruction from CCD samples, IEEE Transactions on Image Processing 8 (1999) 1221–1228.
  7. X. Li, M.T. Orchard, New edge-directed interpolation, IEEE Transactions on Image Processing 10 (2001) 1521– 1527.
  8. X. Li, Demosaicing by successive approximation, IEEE Transactions on Image Processing 14 (3) (2005) 370–379.
  9. R. Lukac, K.N. Plataniotis, D. Hatzinakos, Color image zooming on Bayer pattern, IEEE Transactions on Circuits and Systems for Video Technology 15 (2005) 1475–1492.
  10. F. Malgouyres, F. Guichard, Edge direction preserving image zooming: a mathematical and numerical analysis, SIAM Journal on Numerical Analysis 39 (2001) 1–37.
  11. L. Zhang, X. Wu, Color demosaicking via directional linear minimum mean square-error estimation, IEEE Transactions on Image Processing 14 (2005) 2167–2178.
  12. L. Zhang, X. Wu, An edge guided image interpolation algorithm via directional filtering and data fusion, IEEE Transactions Image Processing 15 (8) (2006) 2226–2238.
  13. Y. Zhu, S.C. Schwartz, M.T. Orchard, Wavelet domain image interpolation via statistical estimation, Proceedings of International Conference on Image Processing 3 (2001) 840–843.