Statistics of Image Neighborhoods

Rather than specifying a region over which to perform a reduction, it is also possible to specify a neighborhood in which to apply a reducer. To reduce image neighborhoods, use image.reduceNeighborhood(). In this case, the reduction will occur in a sliding window over the input image, with the window size and shape specified by an ee.Kernel. The output of reduceNeighborhood() will be another image, with each pixel value representing the output of the reduction in a neighborhood around that pixel in the input image. Figure 1 illustrates this type of reduction.

For example, consider using National Agriculture Imagery Program (NAIP) imagery to quantify landscape differences resulting from logging in the California redwood forests. Specifically, use standard deviation (SD) in a neighborhood to represent the difference in texture between the logged area (SW of the image in Figure 2) and the protected area (NE of the image in Figure 2). For example, to get texture of a NAIP Normalized Difference Vegetation Index (NDVI) image, use reduceNeighborhood() to compute SD in a neighborhood defined by a kernel:

// Define a region in the redwood forest.
var redwoods = ee.Geometry.Rectangle(-124.0665, 41.0739, -123.934, 41.2029);

// Load input NAIP imagery and build a mosaic.
var naipCollection = ee.ImageCollection('USDA/NAIP/DOQQ')
  .filterBounds(redwoods)
  .filterDate('2012-01-01', '2012-12-31');
var naip = naipCollection.mosaic();

// Compute NDVI from the NAIP imagery.
var naipNDVI = naip.normalizedDifference(['N', 'R']);

// Compute standard deviation (SD) as texture of the NDVI.
var texture = naipNDVI.reduceNeighborhood({
  reducer: ee.Reducer.stdDev(),
  kernel: ee.Kernel.circle(7),
});

// Display the results.
Map.centerObject(redwoods, 12);
Map.addLayer(naip, {}, 'NAIP input imagery');
Map.addLayer(naipNDVI, {min: -1, max: 1, palette: ['FF0000', '00FF00']}, 'NDVI');
Map.addLayer(texture, {min: 0, max: 0.3}, 'SD of NDVI');

Any pixel with a non-zero kernel value is included in the computation. The kernel weights are used by default, though you can change that behavior with the inputWeight argument. The input image and reduceNeighborhood() output are compared in Figure 2.