Setting the FFT Filter
The Registration Properties graph shows a power-spectrum histogram
(red- line) where the resolution increases from left to right. The rightmost
part deals with the finest details of the image. Sharp images have a
longer tail (to the right) than blurred images. The FFT filter line (blue)
should be positioned well to the right of the middle of the bend in the
curve, but not to where the red line reaches y=0. Changing the Alignment Filter pixel value
will reposition the blue FFT filter line along the power spectrum.
Press the “Recalc FFT” button and now look at the coloured FFT-spectrum image. The red area
in the middle is where the program estimates the best alignment peak. During processing this
area should be restricted to a small patch. Increasing the FFT alignment filter value and
pressing 'Recalc FFT' will decrease the size of the red area. Decreasing the FFT-filter value
will increase the red area size. Unchecking the alignment filter box and pressing “Recalc
FFT” will display a very small red patch (or even a single pixel). But often you will see signs
of other areas that are also rather high (orange); this can lead to misalignment. Set the
FFT-filter again on and reduce the patch so it’s not too big and not too small. Often you
can keep these settings for many of your images acquired in the same fashion (depending
partly on the alignment box too) the same.
Setting the Quality Estimate Filter Band
This is another filter that works much alike the FFT alignment filter. Using
the power-spectrum in the graph, the relative sum of the data between the
two green lines is calculated as a proportion of the total area below the
graph. Images that are sharper/clearer have a longer right-hand tail in a power-spectrum
histogram. Set the estimate (use the thumbwheel and read the quality value from the label in
the Registration Properties graph) to a value between 0.1 - 0.2. Setting the value higher will
discriminate less between images and setting it too low can result in wrong estimates. Be
aware that the estimation procedure is automated and by no means as good as the human
eye. But during the later processing stages you can still reject lesser images and the
optimization also will improve the alignment further.
21
Summary of Contents for NexImage
Page 14: ...13 User s Guide...
