Hello. Will 'combine classes' work? http://www.harrisgeospatial.com/docs/CombiningClasses.html

Alos, read this: http://www.harrisgeospatial.com/Support/SelfHelpTools/HelpArticles/HelpArticles-Detail/TabId/2718/ArtMID/10220/ArticleID/18580/How-to-change-the-classification-values-pixel-values-in-an-ENVI-Classic-classification-image.aspx

Dear Tharindu,

You can reclassify your images in any GIS software or in R. The operation is called "reclass" and it basically changes the values of a raster/image to a set of other values you provide. For doing so in:

1) ArcGIS: http://desktop.arcgis.com/en/arcmap/10.3/tools/spatial-analyst-toolbox/reclassify.htm

2) GRASS GIS: https://grass.osgeo.org/grass70/manuals/r.reclass.html

3) QGIS: https://gis.stackexchange.com/questions/17712/performing-raster-reclassification-in-qgis

or

4) R: https://www.rdocumentation.org/packages/raster/versions/2.6-7/topics/reclassify

I hope that helps,

Maria

Hi Tharindu,

The easiest way is to use Mosaic tool in ArcGIS (Arctoolbox -> Data Management Tools -> Raster -> Raster Dataset -> Mosaic To New Raster).

Your both classified images will be used as inputs and trick is to use mosaic operator wisely. One way is to code 2nd classified image (shadow classes) with greater numbers and then use "Maximum" option in Mosaic Operator and then combine classes in new raster dataset. Another option (which I feel more appropriate, as you have already isolated shadow areas) is to use 2nd classified image as priority in overlapping area, for example your 2nd image is last in inputs then set Mosaic Operator as "Last". I hope it will work but please mention if you still need further assistance.

Sorry for not providing you solution in ENVI, but you may think on same line and may find similar tool in ENVI as well

Regards

Hello,

You can combine 2 images using a band-to-band operation. ArcGIS has a tool called Raster Calculator; in ENVI, it is called Band Math.

Follow the general steps described here in order to understand how it works: http://www.harrisgeospatial.com/docs/BandMath.html

Load both images in ENVI. Let's say your 2 images are called B1 (1st classification) and B2 (2nd classification), and the shadow class is represented by the value 8 in B1. The expression you have to use in Band Math is:

B1 * (B1 NE 8) + B2 * (B1 EQ 8)

Explanation:

The result you want can be viewed as:

- keep classes from 1st classification if not shadow class: B1 * (B1 NE 8)

- keep classes from 2nd classification if shaded area : B2 * (B1 EQ 8)

(B1 NE 8) and (B1 EQ 8) can be seen as masks where shadow pixels have values of 0 and 1, respectively:

- B1 NE 8 returns an array with 0 for pixels corresponding to shaded area and 1 otherwise; when multiplied with B1, it gives you an array where shadow pixels = 0 and pixels for other classes (from the 1st classification) keep their value.

- B1 EQ 8 returns an array with 1 for pixels corresponding to shaded area and 0 otherwise; when multiplied with B2, it gives you an array where shadow pixels are replaced by pixels from the 2nd classification. Non-shadow pixels take the value 0.

The final addition combines both intermediary results: B1 * (B1 NE 8) + B2 * (B1 EQ 8)

Hopefully, it was clear enough.