MINDIS

Description

Performs multispectral classification of image data for up to 256 classes, based on Euclidean distance. The output is a theme map directed to a database image channel.

Parameters

Name	Type	Length	Value range
InputRaster: Input signature image layers	Raster port	0 -
InputSIG: Input class signature subset 1 *	SIG port	1 - 256	-1024 -
Output: Output raster channel *	Raster port	1 - 1
Mask: Area mask	Bitmap port	0 - 4	Xoffset, Yoffset, Xsize, Ysize
Report	String	0 - 192	See parameter description

* Required parameter

Parameter descriptions

InputRaster: Input signature image layers

Specifies the image layers to be classified.

InputSIG: Input class signature subset 1

Specifies the class signature segments (type 121) to use in the classification.

Up to 256 segments can be handled and up to 16 integer values may be specified.

Output: Output raster channel

Specifies the channel to receive the output theme map.

Mask: Area mask

Specifies the window that defines the area to be processed within the input raster.

The four values specified define the x,y offsets and x,y dimensions of a rectangular window identifying the area to process. Xoffset, Yoffset define the upper-left starting pixel coordinates of the window. Xsize is the number of pixels that define the window width. Ysize is the number of lines that define the window height.

If no value is specified, the entire channel is processed.

Report

Specifies where to direct the generated report.

Available options are:

• LOG: generates a report on the terminal (default)
• <modulename.RPT>: appends a report to the module's RPT file
• DISK: generates a report on file "IMPRPT.LST"

Details

Performs multispectral classification of image data for up to 256 classes based on Euclidean distance. The output is a theme map written to the specified output image channel. The description of the output image channel will be determined by taking the first three characters from the name of each input class signature segment.

The MINDIS classifier differs from MLC, in that it is based on the shortest Euclidean distance to class means, rather than on the likelihood measure employed by MLC.

The execution progress may be monitored.

MINDIS classifies image data on an image file (FILE) using a set of 256 possible class signature segments as specified by DBS1. Each segment specified in DBS1 stores signature data pertaining to a particular class. Only the mean vector in each class signature segment is used; other data, such as standard deviations and covariance matrices, are ignored (this information is used by the maximum likelihood classifier, MLC).

The result of the classification is a theme map written to a specified output image channel (DBOC). A theme map encodes each class with a unique gray level. The gray-level value used to encode a class is specified when the class signature is created (using the parameter VALU for CSG). If the theme map is later transferred to the display, a pseudocolor table should be loaded so that each class is represented by a different color.

The MASK parameter specifies the area to be processed. Pixels outside this area on the output channel will not be modified. Only the area under the mask will be classified; the remainder of the image will not be processed.

The four values specified for Mask define the x,y offsets and x,y dimensions of rectangular window within the image to be classified.

MINDIS generates a classification report. Sub-area totalizations may be obtained by running MLR.

MINDIS assigns each pixel to the class that has the minimum distance between the pixel value and the class mean. Compared to MLC, which takes class covariance matrices into account, MINDIS generally executes more quickly, but may produce poorer classification results.

It is suggested that MINDIS be used when a quick examination of the classification result is required. MINDIS may also be used in applications in which a simple classification is sufficient.

In situations where MLC's multivariate normal distribution assumption does not hold (for example, signature classes generated from texture images which may not have a single-mode normal distribution), MINDIS may perform even better than MLC, because MINDIS does not make assumptions.

Signatures

Signatures are created by CSG (supervised classification). The signature for each class includes the mean vector, standard deviations, and covariance matrix. MINDIS requires only the class mean vector; the remaining data is ignored.

Under some circumstances, you may want to create a class with an explicit user-specified mean vector. In this case, the best approach is to create a dummy signature segment (using CSG and an arbitrary bitmap), then run CSE to modify the resulting signature segment means. The modified signature segment can then be used by MINDIS (although it would be useless for MLC).

Report

An example output listing produced by MINDIS is shown below.

Seg		Segment number of class signature
Name		Name of class signature segment
Code		Segment value (code) of class signature
		(pixel value used to encode theme map)
Pixels		Number of pixels in class
%Image		Percentage of image covered by class

irvine.pix            [S   11PIC     512P     512L]  14-Aug-90

Seg	Name	Code	Pixels	%Image

17	Water1	10		  6955	  2.65
18	Water2	20		  1098	  0.42
19	Urban	30		 29031	 11.07
20	Range	40		115954	 44.23
21	Crop1	50		 40697	 15.52
22	Crop2	60		 27341	 10.43
23	Crop3	70		 4615	  1.76
24	Forest	80		 36453	 13.91

Total				262144	100.00

Algorithm

The equation used by MINDIS is the minimum Euclidian distance classifier defined by the following equation:

                 T
   Gi(X) = (X-Ui) * (X-Ui)

         = SUM[(xj-uj)**2]  for j = 1 to d
         

The distances between the pixel to be classified and each class center are compared. The pixel is assigned to the class whose center is the closest to the pixel.

If for all i not equal j, Gj(X) < Gi(X), then X is classified as j.

The diagram below illustrates the classification of 4 pixels into 3 classes, pixels are classified to the nearest class center.

 |             Class A                    *   class center
 |  .    a                      .        ...  class boundary  
 |     .           *          .
 |        .  b             .               Pixel      Class
 |          .          .                  
 |              .    .                       a           A
 |          c     .                          b           A
 |                .                          c           B
 |          *      .     *  Class C          d           C
 |                 .  d                                          
 |       Class B   .                        
 |                .                             
 |                              
 +----------------------------------

References

Hodgson, M.E. (1988). Reducing the computational requirements of the minimum-distance classifier. Remote Sensing of Environment, Vol. 24.