Im reading up on this article, Image Classification using Random Forests and Ferns. Its written:
SIFT descriptors [19] are computed at points on a regular grid with spacing M pixels. At each grid point the descriptors are computed over four circular support patches with different radii, consequently each point is represented by four SIFT descriptors. Multiple descriptors are computed to allow for scale variation between images.
Im not sure what the part about four circular support patches means. Anyone who can clarify this to me?
The SIFT descriptor takes an image region and computes a vector that is invariant under translation, scaling, and rotation. Your paper's author computes SIFT descriptors for four circular image regions with different radii at each grid point. This will enable him to recognize similar regions in other images that have been translated, scaled, or rotated.
Support is a mathematical term meaning the values for which a function is non-zero. In this case, it means the pixels over which the descriptor is constructed.