307ウェーブレット畳み込み演算によるノイズ処理

抄録

We have studied the visualization of the current distributions from the locally measured magnetic fields. In order to visualize the current distributions, it is intrinsically reduced into solving an ill-posed inverse problem. In most cases, the solutions of the ill-posed inverse problems tend to increase the noise, and the threshold operation is used for the noise reduction. Though this operation is strong noise reduction strategy, it needs a lot of trial and error and experiences for deciding the threshold value.
In the present paper, we propose one of the methodologies of the noise processing. At first, we apply the generalized vector sampled pattern matching (GVSPM in short) method to visualizing the quasi-3D current vector distributions. Second, in order to extract the major current vectors, we apply the 3D vector wavelet transforms employing the several wavelet base function to the noisy current vectors. Consequently, in order to extract common current vectors among them, each of the current vectors was convoluted. Thus, we have succeeded in visualizing the quasi-3D current vector distributions without any threshold operation.