[GIS] Removing a specific band from an image composite with hundreds of bands in Google Earth Engine

google-earth-engine

In Google Earth Engine, the function glcmTexture() outputs 18 types of texture bands. Out of these, one band (suffixed as _maxcorr) is actually not computed by the servers but still exists in the output as a GLCM band with all zero pixel values.

This _maxcorr band creates a problem when band-wise normalization of the image composite is attempted by the unitScale() function as this function requires min value to be less than the max value.

How to tackle this issue?

I thought of two possibilities:

Prevent glcmTexture() from generating the "_maxcorr" band
OR
Remove "_maxcorr" band before input to unitScale() function

How to remove a specific band from an image in GEE?
Trying reverse by using .select() function is not feasible as there are hundreds of other bands in the composite that need to be selected.

Best Answer

In order to remove particular bands from an image, you can take the image's bandNames, filter that list to only the bands you want, and select those bands from the image:

function removeMaxCorr(glcmImage) {
  return glcmImage.select(
    glcmImage.bandNames().filter(
      ee.Filter.stringEndsWith('item', '_maxcorr').not()));
}

Map.addLayer(removeMaxCorr(myImage.glcmTexture()));

(The pseudo-property name 'item' is the name used to refer to the value of a List item that isn't a feature with its own properties — here, a string.).

Related Solutions

[GIS] Removing images based on band threshold using Google Earth Engine

This can be done in two steps. First, you need to turn information stored in images into metadata (image property). Then it can be used in the same way as all other filters. For example, this is how you can filter all images with some number of potential water pixels within some distance from a given point: https://code.earthengine.google.com/491ecb33a4b5fbe5f4cb9a3669254d05

// get all images for a given geometry
images = images.filterBounds(geometry)
print('Number of images: ', images.size())

// estimate potential water probability using MNDWI
var water = images.map(function(i) {
  var water = i.divide(10000).normalizedDifference(['B3', 'B8']).unitScale(-0.05, 0.15)

  return water.mask(water)
})

// estimate number of pixels above some thredhold within 500 m buffer around given point

// define area of interest and scale
var aoi = geometry.buffer(500)
Map.addLayer(aoi)

var scale = Map.getScale()

// applies thresholds and computes new property to be used for filtering
function threshold(i) {
  var count = i.gt(0.5).reduceRegion(ee.Reducer.sum(), aoi, scale).values().get(0)

  return i.set({water_pixel_count: count})
}

water = water.map(threshold)

print('Number of water pixels for all images:', water.aggregate_array('water_pixel_count'))

// filter
water = water.filter(ee.Filter.gt('water_pixel_count', 100))

print('Number if images after filtering: ', water.size())

require('users/gena/packages:animation').animate(water)

The above code computes a new property for every image and then uses that property to filter image collection, resulting in only images where water is present.

[GIS] Extracting pixel values by points from multi-band image and converting to table in Google Earth Engine

For a single (and big) table it's just one more piece of the mamushka:

var mapfunc = function(feat) {
  var geom = feat.geometry()
  var addProp = function(img, f) {
    var newf = ee.Feature(f)
    var date = img.date().format('yyyy-MM-dd')
    var bands = img.bandNames()
    var value = img.select(bands)
    .reduceRegion(ee.Reducer.first(), geom, 30)

    var fun = function(name, ini) {
      var ini = ee.Feature(ini)
      var val = value.get(name)
      var newname = ee.String(date).cat('-').cat(name)
      return ee.Feature(ee.Algorithms.If(val,
                        ini.set(newname, ee.String(val)),
                        ini.set(newname, ee.String('No data'))))
    }

    return ee.Feature(bands.iterate(fun, newf)) 

  }
  var newfeat = ee.Feature(glcm_sub.iterate(addProp, feat))
  return newfeat
};

and for exporting one table per band you have to make some client-side code because Export is a client-side function

var compute_band = function(bandname) {
  var mapfunc = function(feat) {
    var geom = feat.geometry()
    var addProp = function(img, f) {
      var newf = ee.Feature(f)
      var date = img.date().format('yyyy-MM-dd')
      var value = img.reduceRegion(ee.Reducer.first(), geom, 30).get(bandname)
      return ee.Feature(ee.Algorithms.If(value,
                                         newf.set(date, ee.String(value)),
                                         newf.set(date, ee.String('No data'))))
    }
    var newfeat = ee.Feature(glcm_sub.iterate(addProp, feat))
    return newfeat
  };
  return fg_points.map(mapfunc);
}

var bands = glcm_sub.bandNames().getInfo()
for (var band in bands) {
  var name = bands[band]
  var table = compute_band(name)
  // export to drive
  Export.table.toDrive(table,
  "Jam_III_2015_GLCM_16-16-"+name, //job name
  "test_GEE_export", //folder name
  "Jam_III_2015_GLCM_16-16-"+name); //file name
}

Best Answer

Related Solutions

[GIS] Removing images based on band threshold using Google Earth Engine

[GIS] Extracting pixel values by points from multi-band image and converting to table in Google Earth Engine

Related Question