The Use Case:

Scientists need to retrieve readily processed data, rather than doing large downloads and programming extraction and analytics with tedious own programming. Analysis-Ready Data (ARD) is not the end, but the beginning: heavy-lifting analytics needs to be done in the server, from any end device.

The service:

With WCPS, process data up to the complexity of, e.g., a Discrete Fourier Transform (DFT). If this is not enough, rasdaman allows dynamic server-side extension of the language with user-defined operations.

Here we demonstrate convolution, a technique widely used, for example, as a preprocessing step for feature extraction. A filter kernel is a small matrix defining weights according to which new pixels are derived from the original pixel plus some neighbourhood pixels. The kernel used here is the Sobel operator, widely used for edge detection (source: Wikipedia):


A WCPS Sobel query makes use of the coverage constructor, which builds the result matrix, in combination with the condenser, which aggregates the new value of each pixel.

Choose Band for Edge Detection:
Original RGB Band

Sobel-Filtered Image