Object smoothing is a typical step at the end of the analysis process to produce for instance map like results. eCognition offers users several strategies doing this:

• direct on vector objects (i.e. ‘vector smoothing’, ‘vector simplification’, and more)
• or on image objects (i.e. ‘morphology’, ‘pixel based density filter’, ‘pixel-based object resizing’and more).

The possibilities are unlimited to combine the available algorithms to customized smoothing strategies.

A common approach is the “grow and shrink” object reshaping strategy using the surface tension of objects. Customers can execute for instance an algorithm to modify the shape of image objects by adding (first) and removing (second) pixels according to a defined surface tension. Sounds complicated, but it is quite simple.

Christian Weise created (in 2011) a customized algorithm to execute such a “grow and shrink” strategy on image objects to smooth the object borders with the ‘pixel-based object resizing’ algorithm. Users must select just the input objects via the algorithm domain and the strength of the smoothing via the ‘Strength’ parameter (moving window size). Very easy to use, you can check if such a strategy make sense for your project(s).

To use the ‘smooth image objects’ algorithm please download the customized algorithm and load it into your project, after it you will find the ‘smooth image objects’ in the algorithm list. The ‘Example Project - Smooth Image Objects’ gives you an overview how to use the algorithm.

Download ‘Smooth Image Objects’ algorithm

Download 'Example Project - Smooth Image Objects'

Interesting related content:

• eCognition Deconstructed video: Pixel-based Object Resizing I
• eCognition Deconstructed video: Pixel-based Object Resizing II
• eCognition Deconstructed video: Morphology
• Project & Rule Set examples: Pixel-Based Object Resizing Object Generalization Complex Landcover
• Project & Rule Set examples: Example Project - Pixel-Based Object Resizing Surface Tension Artificial
• Project & Rule Set examples: Building Generalization