Gaussian Denoising
Gaussian noise of standard deviation SIGMA=25 was added to the original image (gray levels between 0 and 255) using the MATLAB command +SIGMA*RANDN. The images were then denoised using DMMD’s algorithm and the following three published algorithms:

  1. M. S. Crouse, R. D. Nowak, and R. G. Baraniuk. Wavelet-based statistical signal processing using hidden markov models. IEEE Trans. On Signal Processing, 46(4):886-902, 1998.
  2. Xin Li and Michael T. Orchard. Spatially adaptive image denoising under overcomplete expansion. In IEEE Proc. ICIP., 2000.
  3. S. Grace Chang, B. Yu, and Martin Vetterli. Spatially adaptive wavelet thresholding with context modeling for image denoising. IEEE Trans. Image Processing, 9(9):1522-1531, 2000.
  4. DMMD AQua Denoising.

The results are listed below. From left to right they are: wavelet based, spatially adaptive, spatially adaptive with thresholding, DMMD’s AQua denoising.

Leaves: Noisy and Original

Leaves: Denoised

Salt and Pepper Denoising
Images were scaled between 0 and 1, then 30% salt and pepper noise was added using the Matlab command IMNOISE. DMMD’s denoising algorithm was compared against:

[1] The Median filter.

[2] Xin Li. Edge Directed Statistical Inference with Applications to Image Processing. PhD thesis, Princeton University, Princeton, NJ, 2000.

The results are listed in the following table. From left to right: median, Edge Directed Statistical, and DMMD’s.

Boat: Noisy and Original


Boat: Denoised
X-Ray Denoising
When taking XRays, often times the energy, from the radiation generator, can damage the pixels inside the sensor and produce an effect similar to salt and pepper noise. At DMMD we have developed several algorithms that detect and remove the noisy pixels.