User's Guide

SPECCOMB

Combines Level I and Level II classification images

Function:

Combines the Level I classification and all Level II classifications on images produced by SPECSTRT, into a single multi-level classification.

Parameters:

MASKIMG
Input mask (pattern) image. Must be single band, BYTE data. Usually produced by BAYES, to be used along with the multispectral image to produce output multispectral class images. Should be the same as that used with SPECSTRT.

OUT
Output image. The output classified image will be BYTE and the same size as the input pattern image with the pixel values replaced by those from the Level II classified images after applying a mapping to make them unique.

IN
Input images. Must be single band, BYTE data, with no window. At most, eight classified images may be specified. These are obtained by running SPECSTRT, ISOCLASS, EDITSTAT and BAYES on a multispectral image.

MASKVAL
Mask values. Gray level values in the pattern image that correspond to the classified images in IN.

PRINT(TERM)
Output destination.

  = --       : No Report
  = TERM     : Terminal.  The output is sent to the 
               user's terminal.
  = LP       : Line Printer.  The output is sent to the 
               printer defined by $PRINTER.
  = Filename : User supplied file name.  The output is 
               sent to the user supplied file with the 
               extension ".prt".

Example:

  1. LAS> speccomb maskimg=balt.class.l1 out=balt.class in=(balt.water,balt.forest,balt.agri,balt.land) maskval=(1,2,3,4) print=term

    SPECCOMB replaces the pixels of classes in the input Level I classified image BALT.CLASS.L1 with the corresponding pixels from the respective Level II classified images BALT.WATER, BALT.FOREST, BALT.AGRI, or BALT.LAND and produces a multi-level classified image BALT.CLASS. The output is directed to the line printer.

    If the gray levels in BALT.CLASS.L1 were 1-4, then a typical assignment of gray levels in BALT.CLASS may be as follows:

                                                 OUTPUT GRAY
             INPUT IMAGE    INPUT GRAY            LEVEL IN
                NAME          LEVEL              BALT.CLASS
    
             BALT.CLASS.L1      1                SPLIT ( 5- 6)
             BALT.CLASS.L1      2                SPLIT ( 7-10)
             BALT.CLASS.L1      3                SPLIT (11-13)
             BALT.CLASS.L1      4                SPLIT (14-15)
             BALT.WATER         1-2               5- 6
             BALT.FOREST        1-4               7-10
             BALT.AGRI          1-3              11-13
             BALT.LAND          1-2              14-15
    

Description/Algorithm:

The pattern image is a Level I classified image with a small number of classes and is the same as that input to SPECSTRT. The Level II classified images are usually obtained by classifying the output multispectral images from SPECSTRT. (See User Notes.)

Each gray level (corresponding to a Level I class) in the pattern image is replaced by new unique gray levels obtained by mapping the gray level(s) in the corresponding Level II classified image and is marked "split." If the mapped gray level is > 255, it is set to 255. If the user specified no classified image corresponding to a pattern image gray level, that gray level appears unmapped in the output image. If a specified gray level is not found in the pattern image, the program will inform the user and continue execution.

Note that the gray levels in each classification image are mapped to the next available unique gray level on the pattern image. For example, if the pattern image had gray levels 1-8, and the Level II classified image corresponding to the pattern image gray level 1 had gray levels 1-4, then the the corresponding pixels in the output image will have gray levels 9-12.

Nonfatal Error Messages:

  1. [speccomb-nogray] Specified gray level XXX not found in pattern image

    One of the gray levels specified was not found in the pattern image. The Level II image corresponding to the specified gray level is not used for creating the output image.

  2. [speccomb-outgray] Some output gray levels exceed 255; set to 255

    Some of the gray levels in the output image are mapped to values more than 255. These values are set to 255. To avoid this, rerun with images having smaller number of classes.

Fatal Error Messages:

  1. [speccomb-patbands] Pattern image must be single band

    Pattern image must be the same classified image used in SPECSTRT.

  2. [speccomb-pattype] Pattern image must be a BYTE image

    Pattern image must be the same classified image used in SPECSTRT.

  3. [speccomb-inbands] Level II classified image(s) must be single band with no window

    Level II images must be single band classified images with no windows.

  4. [speccomb-intype] Level II classified image(s) must be BYTE

    Level II images must be classified images of BYTE data type.

  5. [speccomb-size] Level II Classified image and Pattern image sizes do not match

    Number of pixels per line in Level II classified images must match number of pixels per line in Level I pattern image.

  6. [speccomb-maskval] Number of gray levels and number of images do not match

    Specify one gray level for each Level II image.

  7. [speccomb-sameval] Same gray level values specified for different images

    Specify a different gray level for each Level II image.

User Notes:

  1. A typical processing scenario is as follows:

        o  Run ISOCLASS to obtain broad Level I classification
           statistics.
     
        o  Run EDITSTAT to reduce and combine level statistics to a
           smaller number of classes, e.g., four groups, WATER, 
           LAND, AGRI, and FOREST.
     
        o  Run BAYES to categorize the entire area into Level I
           classes.
     
    
    Run SPECSTRT to produce a separate multispectral image for each of the classes, e.g., four multispectral images are produced containing all the data in the original multispectral image. Repeat the following on each class image produced by SPECSTRT:

        o  Run ISOCLASS on the image.
     
        o  Run EDITSTAT to reduce and combine statistics from
           ISOCLASS to Level II.
     
        o  Run BAYES on the image to produce Level II categories.
     
        o  Run SPECCOMB to recombine and map the Level I and all 
           Level II classifications into a single multilevel 
           classification.