Name
Abstract | ||
|---|---|---|
1 Abstract Wavelet-based denoising methods have the advantage over low-pass filtering in that relevant detail information is retained, while small details due to noise are discarded. This paper reports a novel technique of removing noise, using a wiener filter in wavelet domain, from an fMRI data and display selectively event related voxels in a spatial domain. After the fMRI signal is wavelet transformed to temporal domain, its median and mean become symmetric as described in Section and its median of absolute deviation (MAD) is calibrated with the standard deviation of a Gaussian distribution to approximately determine the standard deviation of the noise in the fMRI signal. Once the noise level is determined, the result is used to determine the power spectrum of the coherent signal and the corresponding standard variable. The whole process is used to accurately estimate the power of the coherent signal and the associated noise at a given voxel location. The estimated power spectrum is used to approximate the optimal wiener filter coefficient for removing the noise from the fMRI signal. In this experiment, only the signal to be restored is available and all prior knowledge about the ideal signal has to be estimated from it. Though an fMRI is contaminated by both Gaussian and Rician distribution related noises, it is safe to assume noise in an fMRI signal to be additive white Gaussian noise. In this paper, we showed that the power spectrum estimated from this single copy of degraded signal is a true power spectrum of the signal, and as a result, the restoration filter or noise removing filter is optimal, though there is a lack of accurate prior information. The method successfully removes noises and exposes activated voxels in fMRI signal all the time as shown in this paper. |
Year | Venue | Keywords |
|---|---|---|
2006 | IPCV | standard deviation,gaussian distribution,wavelet transform,additive white gaussian noise,power spectrum,low pass filter,wiener filter |
Field | DocType | Citations |
Gaussian filter,Wiener filter,Pattern recognition,Wiener deconvolution,Second-generation wavelet transform,Artificial intelligence,Discrete wavelet transform,Gaussian noise,Wavelet packet decomposition,Mathematics,Wavelet | Conference | 1 |
PageRank | References | Authors |
0.40 | 5 | 4 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Debebe Asefa | 1 | 1 | 1.08 |
| Dinesh P. Mital | 2 | 107 | 32.92 |
| Syed Haque | 3 | 23 | 11.48 |
| Shankar Srinivasan | 4 | 22 | 15.15 |