Abstract: In traditional Spatially Selective Noise Filtration (SSNF), the correlation is weakened by wavelet coefficient shift among different decomposition scales, and the wavelet coefficients are seriously affected by the noise at fine scales. To resolve these problems an improved Spatially Selective Noise Filtration (ISSNF) algorithm is proposed. The relative shift of wavelet coefficients between two adjacent scales is deduced, and then a shift correlation algorithm is proposed. Considering that the higher the scale is, the stronger the correlation of wavelet coefficient is, and that divergence happens during decomposition, a reverse processing method is proposed, in which constraints are provided from higher scale to lower scale. According to statistic principle, the wavelet coefficient of the highest is processed by threshold filtering. The performances of the proposed algorithm, the traditional algorithm and two existing improved algorithms are compared by processing Blocks signal and Heavy Sine signal. Moreover, the above filtration algorithms are applied in group delay estimation of certain GEO satellite. Results of Monte Carlo simulation and processing of measured data demonstrate that the performance of the proposed algorithm is better in noise reduction.

Key words: information processing technology, ssnf, shift correlation, reverse order processing, threshold filtering, group delay estimation