buttons-03.png

Custom Mosaics

Contributed by CEOS SEO/AMA

A mosaic is a composite image created by combining the most appropriate pixels from a collection of ​source images. A common use case is to create cloud-free images for applications that are not time-dependent. There are a number of approaches.

Real Pixel Mosaics

These are cloud-filtered mosaics based on time series data where a specific non-cloudy pixel is selected from a time series. The resulting product will be made up of pixels from different times, and the overall composite may have varying cloud cover percentage throughout, as these algorithms only consider the time series for each individual pixel, meaning cloud cover percentage may vary between pixels. Common selection techniques include:

Most Recent / Least Recent Pixel 

A cloud-free mosaic using the most recent (newest) or least recent (oldest) clear pixel in a time series.

Maximum Value Composite (MVC)

A cloud-free mosaic based on the maximum value of an index (e.g. NDVI) in a time series. This type of mosaic is best for agriculture or forest detection in a time series. In some cases, the minimum value composite is also desired.

Synthetic Pixel Mosaics

These are cloud-filtered mosaics produced using an algorithm to determine the desired spectral combination. Two common algorithms are:

Median Mosaic

Creates cloud-filtered mosaics using the "middle" spectral response of each band in a time series. Since the bands can vary independently, the result is often a "synthetic" product that is not representative of any particular time. This type of mosaic is also possible with radar data.

 

Geomedian Mosaic

A new algorithm developed by Geoscience Australia (GA) that creates a synthetic mosaic, similar to a median, but which is more representative of the spectral relationship between bands.

Explore the Application

View the Notebook and sample outputs here

Want to use the example Notebook on your own Data Cube and data? 

Download the Notebook from GitHub here and run it on your own Cube in a Box installation.

 

 

Tutorial

 

 

 

 

 

 

 

 

 

The video provides a run through of the example Notebook linked above, with the steps listed here:

  1. All DCAL Notebooks begin by specifying a location, time window, and product of interest.

  2. The extents and products are examined and selected in the first 6 cells of the notebook, with user options available in 3rd and 5th cells, marked by "# CHANGE HERE >>>>>>>>>>>>>>>>>>".

  3. Any values within the bounds detailed in the 4th cell will work, and it is suggested to keep the region and time period small. The resulting area is shown in red in cell 6.

  4. You can click on the map to query the coordinates under the cursor.

  5. In cell 7, you load a region of the Data Cube using the selected extents, products, and measurements. The resulting data are summarised in cell 8.

  6. The Median Mosaic is generated in cell 13.

  7. The Geomedian Mosaic is generated in cell 15.

  8. The Most Recent Pixel Mosaic is generated in cell 17.

  9. The Maximum Value Composite (NDVI) is generated in cell 19.

  10. Finally, the resulting mosaics can be exported as a GeoTIFF file for download.