• Noisy sensor data

We propose new approaches to demosaicking of spatially sampled image data observed through a color filter array, in which the correlation of color components are exploited in order to reconstruct a subsampled image. Two frameworks for applying existing image denoising algorithms to color filter array data are developed—one for wavelet domain processing (PSDD), and the other for pixel domain processing (TLSD).

Posterior Sparisty-Directed Demosaicking

Posterior Sparisty-Directed Demosaicking (PSDD) is a wavelet-based demosaicking method (Korneliussen 2014, Hirakawa 2007).  It was introduced as a part of “camera pipeline for chromatic aberration” (Korneliussen 2014). Chromatic aberration (CA) breaks most demosaicking methods => introduces zippering.  PSDD tolerates CA by doing away with the cross-color correlation assumption, and it is also robust to aliasing and handles noise, textures, and saturated colors.

PSDD is an updated version of our earlier wavelet-based demosaicking method (Hirakawa 2007). As such, PSDD can be combined with any wavelet-based denoising algorithm. In our implementation below, we combined with our Skellam denoising method (designed specifically for camera noise). But you can replace it with your favorite algorithm. PSDD can also yield denoised sensor data (ie CFA sampled data without demosaicking).

  • [Reference Code]
  • If using just demosaicking, please cite Korneliussen 2014.
  • If using combined demosaicking and denoising, please cite as “joint demosaicking and denoising method of Hirakawa 2007, updated with Korneliussen 2014.”
  • If using only CFA denoising, please cite Hirakawa 2007.
  • If using the default Skellam denoising method included in the implementation, please cite also Hirakawa 2009.
  • The full “camera processing with chromatic aberration” implementation, including PSDD, can be found here.
Korneliussen, Jan Tore; Hirakawa, Keigo (2014): Camera Processing With Chromatic Aberration. In: IEEE Transactions on Image Processing, 23 (10), pp. 4539-4552, 2014. (Type: Journal Article | Links | BibTeX)
Hirakawa,; Wolfe, (2009): Skellam shrinkage: Wavelet-based intensity estimation for inhomogeneous Poisson data. In: Information Theory, IEEE Transactions on, (99), pp. 1–1, 2009. (Type: Journal Article | Links | BibTeX)
Hirakawa, (2008): Color filter array image analysis for joint denoising and demosaicking. In: Lukac, Rastislav (Ed.): Single-Sensor Imaging: Methods and Applications for Digital Cameras, pp. 239–261, 2008. (Type: Incollection | BibTeX)
Hirakawa,; Meng, Xiao-Li; Wolfe, (2007): A Framework for wavelet-Based Analysis and Processing of Color Filter Array Images with Applications to Denoising and Demosaicing. In: Acoustics, Speech and Signal Processing, 2007. ICASSP 2007. IEEE International Conference on, pp. I-597 -I-600, 2007. (Type: Inproceeding | Links | BibTeX)

Total Least Squares Demosaicking

Total Least Squares Demosaicking (TLSD) is a pixel domain demosaicking method (Hirakawa 2006 TLS).  It combines signal-dependent image denoising method (Hirakawa 2005) with demosaicking.
[Reference Code]

Hirakawa,; Parks, (2006): Image denoising using total least squares. In: Image Processing, IEEE Transactions on, 15 (9), pp. 2730–2742, 2006. (Type: Journal Article | Links | BibTeX)
Hirakawa,; Parks, (2006): Joint demosaicing and denoising. In: Image Processing, IEEE Transactions on, 15 (8), pp. 2146–2157, 2006. (Type: Journal Article | Links | BibTeX)
Hirakawa,; Parks, (2005): Image Denoising for Signal-Dependent Noise. In: Acoustics, Speech, and Signal Processing, 2005. Proceedings. (ICASSP '05). IEEE International Conference on, pp. 29 - 32, 2005. (Type: Inproceeding | Links | BibTeX)
Hirakawa,; Parks, (2005): Joint demosaicing and denoising. In: Image Processing, 2005. ICIP 2005. IEEE International Conference on, pp. III - 309-12, 2005. (Type: Inproceeding | BibTeX)

Leave a Reply

Your email address will not be published. Required fields are marked *