next up previous contents
Next: Hamming filter Up: Implementation Previous: Method: porbits   Contents

Method: adaptive

After the resampling of the slave on the master grid is performed this algorithm can be used. The local fringe frequency is estimated using peak analysis of the power of the spectrum of the complex interferogram. The resampling is required since the local fringe frequency is estimated from the interferogram. This fringe frequency is directly related to the spectral shift in range direction. (Note this shift is not a shift, but different frequencies are mapped on places with this shift...) The algorithm generally works as follows.

In practice this is done in blocks. These blocks are overlapping in lines (because the averaging over lines means one cannot filter all lines in the block), and not in range. Parameters that can be adjusted are the FFTlength, the moving average mean, the SNR threshold.

The fftlength should be large enough to yield a good estimate of the local fringe frequency, and small enough to contain a constant slope of the terrain. The total number of fringes in range directorion can be easily estimated using the perpendicular baseline.

It is probably a good idea to add a card so an overlap in range between blocks can be used. This avoids 'edge' effects, and increases the filtering of terrain near, e.g., a lake (since the SNR for peak detection will be higher for a number of blocks towards the noise). This is not implemented yet.

See also [5], [7], [3]. See also our matlab toolbox.

Figure 22.2: Peak estimation in spectral domain of (oversampled) complex interferogram. Non FFTshifted.

Figure 22.3: Spectral filtering windows (inverse hamming, boxcar (rect), and new hamming. Note these are FFTshifted.

Figure 22.4: Detail of interferogram with and without rangefiltering. (fftlength=128, nlmean=15, snrtreshold=5). The perpendicular baseline is about 200 m for this interferogram. The fringes are clearly sharper after the filtering. The number of residues for the interferogram was reduced by 20%. Subtraction of both interferograms yielded a random phase, so no structural effect of range filter implemenation is suspected.
\begin{figure}\vbox{ \epsfig{file=Figures/,width=.9\linewidth}
\epsfig{file=Figures/,width=.9\linewidth} }

next up previous contents
Next: Hamming filter Up: Implementation Previous: Method: porbits   Contents
Leijen 2009-04-14