European Commission logo
INSPIRE Community Forum

The Datacube Manifesto

  • Iurie MAXIM
    Iurie MAXIM


    Maybe it should be taken into consideration that an aerial image in visible spectrum is represented usually as RGB (red, green, blue), therefore in one grid point, from the point of view of a database there are 3 values, one for each color. This is already 5 dimensions: X,Y,R,G,B. For satellite images there are usually multiple bands and even they have different resolutions. For example pancromatic band will cover all grid points, but band 3 will has a smaller resolution and NIR band even has much smaller resolution. So therefore not all grid points will have data, if they will be part of the same datacube. None of these aspects are not covered by the paper.

  • Peter BAUMANN
    Peter BAUMANN

    Your post is leading into interesting conceptual discussions wrt dimensionality. A coverage (as well as an image) is a function f: CoordinateSpace -> ValueSpace, such as f1: int, int -> int, int, int. Now you can successively, without loss of information, move values into coordinates ending up with fx: int, int, int, int, int -> boolean - which tells you whether at a particular position there is a pixel with a particular color. Not very storage efficient, though. On a side track, some people really have tried to model imagery as relational tables with a 5D schema as you indicate. MonetDB has tried this and failed, SAP HANA has tried it and is also not happy. Well, arrays are different from sets so technology supporting it well will be different also - "no one size fits all". Sorry for the digression.

    You have made another interesting point: in hyperspectral imagery, not all bands will likely have the same resolution. Now, a coverage in order to be "analysis ready" = simple to deal with needs to be homogeneous in coordinate system and resolution. The new OGC Coverage Implementation Schema 1.1 [1], therefore, has paved the way for coverage collections. These are containers of coverages bundled together. A next step for OGC will be to define functionality on these coverage containers, but that needs some careful design.

    Outing myself: I am the writer of the Manifesto. Actually, this had to be short and focused, it was intended to clarify basics as these days virtually everything tends to be called a datacube. Fully agreed, there is so much more to say!

    thanks for your comment,



Elevation, Ortho & Grids

Elevation, Ortho & Grids

INSPIRE Thematic Cluster Elevation, Orthoimagery, Reference systems, Geographical grids - Join this group to share your knowkledge, learn and collaborate in solving issues related to the Elevation, Orthoimagery, Reference systems and Geographical grids themes