Image Processing

A popup window will open which offers some basic image processing options (see Fig. 5.16).

Figure 5.16: The Image Processing popup window

Boxcar Smooth

will apply a boxcar averaging to the image or spectrum, i.e. each pixel is replaced by the average of its neighborhood. The width is specified in scan pixels (which are different from the displayed pixels if you are on a zoom stage different from 1.0 and if the True Aspect Ratio box is checked). Check the IDL help on SMOOTH for more information.

Median Smooth

will replace each pixel by the Median of its neighborhood. Again, the width is specified in Scan Pixels. Check the IDL help on MEDIAN for more information.

Gaussian Smooth

will replace each pixel by a Gaussian-weighted average of its neighborhood. The width (1/e with of the Gaussian) is specified in scan pixels.

Gaussian Smooth is usually preferred over Boxcar smooth, because its gives greater weight to closer neighbors. Mathematically, a Gaussian smooth corresponds to a Gaussian filter in Fourier space, whereas a Boxcar smooth corresponds to a Sinc filter in Fourier space (which has unwanted side lobes).

Unsharp Masking (Sharpening)

(not for spectra) will apply an Unsharp Mask filter to the image. Check the IDL help on UNSHARP_MASK for more information. The parameters are currently hard coded to the IDL defaults.

Despike

(not for spectra) will despike the image. Ask Chris for details about the algorithm.

Noise Filtering

(not for spectra) is meant for filtering out periodic frequencies. It may be useful for getting rid of non-random line noise for example.

First choose, if you want to have a 1D power spectrum displayed, which shows the frequencies present in all scan lines added up incoherently, or the 2D fourier transform of your image. You can then drag the mouse over the frequency area to be filtered out and choose which type of filter to use. You can either Smooth out a spike (1D only), Kill it, or apply a second order Butterworth (BW) or a Gaussian filter of chosen width. In the 2D case, you can have the fourier transform overlayed with the corresponding time frequencies (calculated over the pixel dwell), by hitting Show Frequency. Don't forget to Apply Filter.

Crop

(not for spectra) lets you crop the image. The button is only sensitive if you have marked a rectangle on your image.

Undo / Redo

lets you undo and redo image processing steps. The number of steps is unlimited (programmatically), so if you overdo it, you might run out of memory at some point. These buttons un/redo only image processing, and nothing else!

Original Image

lets you undo all image processing steps and recover the original image. This can't be undone!

Close

deserves no further explanation.

Some general remarks about image processing:

• None of these actually change your scan data, they will only change the display. The processed data can't be written back to a STXM data file. I.e., the only way to preserve your image processing steps is to write out image files (PNG, JPEG, TIFF, BMP, EPS) or spectra files (CSV, XAS). Binary files will also work.

• You can do the following while preserving your image processing steps:
  • Adjust Min, Max, Gamma

  • Change the colortable

  • Zoom and toggle the True Aspect Ratio setting

  • Set the current spectrum as I_0

  • Print and output files

  But if you do anything which recalculates the displayed image from the raw data, you will lose all image processing. This includes:

  • Toggling the Show I/I_0 setting

  • Changing signal type or detector / display

  • and of course opening a new file

More image processing tools, like Fourier Filtering or Wiener Filtering, might be included in the future if users are interested.