Hugin stacker

From PanoTools.org Wiki
Jump to navigation Jump to search

hugin_stacker stacks overlapping images to a single image.

The general usage is

 hugin_stacker [options] --mode=STRING images

There are several stack modes (switch --mode) available:

  • min|minimum|darkest: Select the darkest pixel
  • max|maximum|brightest: Select the brightest pixel
  • avg|average|mean: Calculate the mean for each position
  • median: Calculate the median for each position
  • winsor: Calculate the Winsor trimmed mean for each position. The parameter can be set with --winsor-trim=NUMBER (default: 0.2)
  • sigma: Calculate the sigma clipped mean for each position. Fine-tune with --max-sigma=NUMBER (default: 2) and --max-iterations=NUMBER (default: 5)

This is useful for e.g.

  • automatic tourist removal (not only tourists, also other moving objects ;-))
  • noise reduction
  • visualize movement
  • multiplicity

Further parameters

  • --output=FILE: Set the filename for the output file (if not given final.tif is used).
  • --compression=value: Set the compression of the output files
    • For jpeg files use values between 0 and 100
    • For tiff files valid values are: PACKBITS, DEFLATE, LZW

Mask input images

Beside the stacked output hugin_stacker can also mask the input images (available only for stacking modes median|winsor|clip). This mode is activated with --mask-input. In this case the stacked image is calculated first. Then each pixel in each image is checked: if the value of this pixel differs more then mask sigma * standard deviation from the mean/median, this pixel is made visible. If it is in the mentioned range the pixel is masked out.

The sigma parameter for this step can be changed with --mask-sigma=NUMBER (default is 2).

You can output a separate mask image for each input image. Set the suffix with --mask-suffix=STRING (default is _mask). Or with the option --multi-layer-output you can output a layered TIFF with the name specified with --output. The file contains the averaged image as layer 0 and all input images as additional layers with the mask as described above.