home left right up

---

Fourier Transform Experimentation

HIPR Applet Running Instructions

The following are instructions on how to run the applets which are part of the HIPR package. These applets are designed to allow the user to try out the operators which are outlined in the worksheets.

You may want to review the general instructions first, which gives a demonstration overview, JAVA details, browser details, known bugs, etc.

Source for this operator.

Image Loading

You can experiment with different images using any of the images supplied in HIPR or by a URL. Some of those available in HIPR have the prefixes given here and .gif as the file extension: grayscale (cam1, che1, cln1, fru1, man8, str1, str2, str3, tol1, txt2, wal2, wom1, wom2, urb1, xra1) and range (bae1, fac1, phn2, ren1, ren2, ufo1, ufo2, ufo3) images.

Enter image specifiers (into the edit box at the top of the screen) either as a full HIPR image filename (eg. cam1.gif) or by a full URL. Only gif files can be loaded. Because of JAVA security, the URL should refer to a publically accessible place, such as your public_html directory.

Pushing the Load Image button or pushing "Return" on the keyboard causes the specified image to be loaded. The input image is displayed in the Input display area below the control buttons, along with the image size.

Note: you should only use square images (i.e. NxN) otherwise this implementation may give erroneous results.

Operator Instructions

The Fourier transform can be applied by clicking on the "Fourier Transform" button. This must always be done first.

The result of the Fourier transform is displayed at the upper right of the panel. There are several outputs that can be displayed, and several ways to display them:

  1. magnitude of the complex number output
  2. logarithm of the magnitude of the complex number output
  3. phase angle of the complex number output
  4. real component of the complex number output
  5. logarithm of the real component of the complex number output
  6. imaginary component of the complex number output
  7. logarithm of the imaginary component of the complex number output
Clicking on the label selects the given image. The Magnitude Log image is often most helpful for understanding the Fourier output and filtering.

Various filtering operations can then be applied to the output of the Fourier transform:

  1. Frequency domain area filter: the four numbers define a box with diagonal from (X1,Y1) to (X2,Y2). Clicking "Remove Area" or "Keep Area" modifies the Fourier transform output.

  2. Apply Notch Filter: this removes pure vertical and horizontal high spatial frequency components. The Width parameter controls the notch width and the Radius parameter controls where the filter starts.

  3. Apply Frequency Filter: This applies a high or low pass stop filter defined radially about the origin. The Radius parameter box defines the cutoff frequency of the filter. The pull down panel selects a High Pass or Low Pass Filter of the given radius.

  4. Gaussian Smooth: the Fourier domain gets a low-pass smoothing filter of the given theta. The Kernel Size controls how large of a mask size is used, and Theta controls the standard deviation of the Gaussian smoothing function.

When a filter is applied, you can see the results in the Fourier domain output at the right. Applying a filter might take a few seconds.

Clicking the "Reset Images" button removes the effects of the filters and allows you to investigate new filters.

The inverse Fourier transform can then be applied to view the effects of the filtering in the spatial domain. The user must click on the "Inverse Fourier Transform" button to do this.

The time box shows the amount of time which the operator took to complete the process on the input image.

Image value inspection

You can inspect the input and result image pixel values by depressing the left mouse button and moving the cursor over the desired pixel. The pixel coordinates and gray-level value are displayed underneath the image. Note that some images are larger than the displayed 256x256 display window. Larger images are subsampled in an appropriate manner.

Known Bugs

You can find out about known problems with the system here.

---

home left right up

©2003 R. Fisher, S. Perkins, A. Walker and E. Wolfart.

Valid HTML 4.0!