Reconstruct - Filtered Backprojection

This command displays a dialog to set the parameters for reconstructing an image from projections using the filtered backprojection technique. The parameters available are:

Filter Selects the filter to apply to each projection. To properly reconstruct an image, this filter should consist of the the absolute value of distance from zero frequency optionally multiplied by a smoothing filter. The optimal filters to use are:

abs_bandlimit
abs_hamming
abs_hanning
abs_cosine

Hamming parameter Sets the alpha level for Hamming window. This parameter adjusts the smoothing of the Hamming filter and can range from 0 to 1. At a setting of 1, the Hamming filter is the same as the bandlimit filter. At a setting of 0.54, the Hamming filter is the same as the Hanning window.

Filter Method Selects the filtering method. For large numbers of detectors, the FFT-based filters are preferred whereas for smaller numbers of detectors convolution can be faster. When Advanced Options have been turned off, this menu only shows the two basic choices: convolution and FFT. However, when Advanced Options have been turned on, additional selections are available as discussed in the next section.

Interpolation Interpolation technique during backprojection. cubic has optimal quality when the data is smooth. Smooth data is obtained by taking many projections and/or using a smoothing filter. In the absence of smooth data, linear gives better results and is also many times faster than cubic interpolation.

nearest - No interpolation, selects nearest point.
linear - Uses fast straight line interpolation.
cubic - Uses cubic interpolating polynomial.

Advanced Options

These options are visible only if Advanced Options has been selected in the File - Preferences dialog. These parameters default to optimal settings and don't need to be adjusted except by expert users.

Filter Method Selects the filtering method. The general comments about this parameter given the previous section still apply. With Advanced Options on, the full set of filter methods are available:

convolution
fourier - Uses simple Fourier transform.
fourier-table - Optimizes Fourier transform by precalculating trigometric functions.
fftw - Uses complex-valued Fourier transform with the fftw library.
rfftw - Uses optimized real/half-complex Fourier transform.

Backprojection Selects the backprojection technique. A setting of idiff is optimal.

trig - Use trigometric functions at each image point.
table - Use precalculated trigometric tables.
diff - Use difference method to iterate within image.
idiff - Use integer iteration techique.

Filter Generation Selects the filter generation. With convolution, direct is the proper method to select. With any of the frequency methods, inverse-fourier is the best method.

direct
inverse-fourier

Zeropad Zeropad factor when using frequency-based filtering. A setting of 1 is optimal whereas a setting of 0 disables zero padding. Settings greater than 1 perform larger amounts of zero padding but without any significant benefit.

ROI These four settings control the region of interest for the reconstruction. The default values match the dimensions of the entire phantom. By constraining the ROI to be a smaller square, the reconstruction will be magnified.

Filter	Selects the filter to apply to each projection. To properly reconstruct an image, this filter should consist of the the absolute value of distance from zero frequency optionally multiplied by a smoothing filter. The optimal filters to use are: `abs_bandlimit` `abs_hamming` `abs_hanning` `abs_cosine`
Hamming parameter	Sets the alpha level for Hamming window. This parameter adjusts the smoothing of the Hamming filter and can range from `0` to `1`. At a setting of `1`, the Hamming filter is the same as the bandlimit filter. At a setting of `0.54`, the Hamming filter is the same as the Hanning window.
Filter Method	Selects the filtering method. For large numbers of detectors, the FFT-based filters are preferred whereas for smaller numbers of detectors `convolution` can be faster. When Advanced Options have been turned off, this menu only shows the two basic choices: `convolution` and `FFT`. However, when Advanced Options have been turned on, additional selections are available as discussed in the next section.
Interpolation	Interpolation technique during backprojection. `cubic` has optimal quality when the data is smooth. Smooth data is obtained by taking many projections and/or using a smoothing filter. In the absence of smooth data, `linear` gives better results and is also many times faster than cubic interpolation. `nearest` - No interpolation, selects nearest point. `linear` - Uses fast straight line interpolation. `cubic` - Uses cubic interpolating polynomial.

Filter Method	Selects the filtering method. The general comments about this parameter given the previous section still apply. With Advanced Options on, the full set of filter methods are available: `convolution` `fourier` - Uses simple Fourier transform. `fourier-table` - Optimizes Fourier transform by precalculating trigometric functions. `fftw` - Uses complex-valued Fourier transform with the fftw library. `rfftw` - Uses optimized real/half-complex Fourier transform.
Backprojection	Selects the backprojection technique. A setting of `idiff` is optimal. `trig` - Use trigometric functions at each image point. `table` - Use precalculated trigometric tables. `diff` - Use difference method to iterate within image. `idiff` - Use integer iteration techique.
Filter Generation	Selects the filter generation. With convolution, `direct` is the proper method to select. With any of the frequency methods, `inverse-fourier` is the best method. `direct` `inverse-fourier`
Zeropad	Zeropad factor when using frequency-based filtering. A setting of `1` is optimal whereas a setting of `0` disables zero padding. Settings greater than `1` perform larger amounts of zero padding but without any significant benefit.
ROI	These four settings control the region of interest for the reconstruction. The default values match the dimensions of the entire phantom. By constraining the ROI to be a smaller square, the reconstruction will be magnified.