We look for a rotation invariant pattern detector for vector fields.
Moment invariants are powerful function descriptors. They have been well studied and often used on scalar fields because they are robust, intuitive, and able to represent objects independent from their specific position, orientation, and scale. Their extension to vector valued functions forms a valuable foundation for several flow visualization tasks, like pattern detection, feature categorization, and clustering.
Finding patterns independent from their orientation.
The algorithm’s response for finding the vortex saddle combination. The brighter the circles, the higher the similarity.