Wavelet de-noising of 3-D SIMS images

To demonstrate the result of the wavelet shrinkage algorithm, several examples with different measurement time, i.e. SNR, were measured. In Fig. 5 an Al distribution is displayed. More examples can be found in [8]. 

For assessing the performance of the above-described wavelet de-noising algorithm a quantitative evaluation of the reconstruction was carried out. As figures of merit, the MSE (Eq. (6)) and the SNR (Eq. (8)) were used. Wavelet de-noising was compared with the optimal MSE Wiener filter [2]. Wiener filter reconstructions were calculated using the wiener function from the 

5 keV scanning steps 512 x 512 step width 0.1 pm analytical area 51.2 x 51.2 pm measurement time per pixel I ms measurement time per image 256 s detected 

MATLAB Image Processing Toolbox [13]. The block size of the Wiener filter was tuned to find the least MSE reconstruction. Since quantification requires true images for comparative reasons, the evaluation was carried out on the basis of a simulated image (Fig. 6), which has simple features, such as rectangular bumps with increasing widths, resembling structures in some real 

SIMS images. A simulated image with Poisson statistics and an SNR = 3.8 was created. The Anscombe transform was applied prior to filtering. 

---