Image Filtering

Learn more about Filtering in SPRING

The Filtering techniques over digital images in SPRING can be accessed at the main menu: Image - Filtering...
The prerequisites for Image Filtering are:

  • 1- having a monochromatic image (in gray levels) and,
  • 2- activating this image, not necessarily presenting it on the screen.

NOTE: The steps to perform image filtering are similar for all types (linear, non-linear or radar). Depending on the type chosen the interface changes, adjusting to your choice. The procedure for editing and executing a mask of a new filter and is similar, but you will have to edit the mask before.

Execute Filtering:

  • activate an Image Infolayer on the "Control Panel" ( );
  • click on Image on the main menu and choose Filtering...;
  • choose the Type (Linear, Non Linear or Radar). For each type chosen, a group of filters is presented Check the table below to see the available options;
  • type in the Name (output layer) on the text box, the IL name to be generated (maximum of 32 characters);
  • in case you want to define an area smaller than the project's, click on Bounding Box... (see more on bounding box);
  • define the Number of iterations, to define how many times the same filter will be applied on the same image;
  • click on Apply .
Types of Filters Options Masks
Linear
Low-Mid Pass 3x3, 5x5 or 7x7
Edge Directional NW, W, SW, N, S, NE, E or SE
Edge Non-Directional Low, Average or High
TM Enhancement Predefined by the system (4x4)
Masks Masks editor
Non Linear
Edge Detection Sobel or Roberts
Morphologic Dilation Mtot, Mx, M+, M-, M, Md or Me
Morphologic Median
Morphologic Erosion
Mask Elements Editor
Radar
Lee Define Type, Window (3x3, 5x5, 7x7, 9x9), Image and Number of Looks.
Kuan
Frost Define Type, Window (3x3, 5x5, 7x7, 9x9), Image, Number of Looks and Correlation Coefficient.

Note: the sequence of operations for executing the Filtering is similar if the mask is defined by the user. You need to select a previously created mask and execute it on the chosen image. Check next how to define a mask.


Masks Editor

In SPRING the user can create his own filters. A mask editor allows to define them for Linear and Non linear filter types.

    • For linear filters the user has the option to create mask with several dimensions using the values 1,3,5 and 7, for instance, 1x3, 3x3, 7x5 etc...
    • For non-linear filters the user has the options Dilation, Median and Erosion, using a 3x3 mask for the weights for the structuring element.

Creating a linear filter mask:

  • click on Image - Filtering... on the main menu;
  • click on Linear to choose the type;
  • choose the option Masks at Linear Filters. Notice that a text box is presented so you can inform the mask name or you can use the button Select... ;
  • click on Select... to create a new mask or choose an existing one from the list. The window "Filtering" is presented;
  • after creating or updating a mask, you will be able to Execute Filtering (the last four steps).

Creating a non linear filter mask:

  • click on Image - Filtering... on the main menu;
  • click on Non Linear to choose the type;
  • choose the option Masks at Non Linear Filters. Notice that a text box is presented so you can inform the mask name or you can use the button Select... ;
  • click on Select... to create a new mask. The window "Structuring Element" is presented;
  • after creating or updating a mask, you will be able to Execute Filtering (the last four steps).

Filtering Radar Images

When you select the Radar type, the interface is changed and the parameters are predefined (you will not be able to create a mask).

Executing Radar Filtering;

  • click on Radar to choose the type;
  • choose the Speckle Noise Filters: Lee, Kuan or Frost;
  • define the parameters in Type (Fixed or Adaptative), Window (3x3, 5x4, 7x7 or 9x9), Image (Amplitude or Intensity) and Number of Looks;
  • define a name for the Image Output;
  • change the Bounding Box... if you wish;
  • click on Apply .

NOTE: The number of Looks represents how many "looks" the image was generated and this information is usually stored on the header of the image (see "Multi-Look Processing"). For the Frost filter you should define the correlation Coefficient. (LAG =1, indicates autocorrelation on a pixel neighborhood). This value can be calculated at the SPRING statistical module.

See also:

About RADAR Images
How to obtain information about Filtering techniques
Main Menu - Image