Ordinal Coding of Image Microstructure
Alexey Koloydenko and Donald Geman
Eurandom, Eindhoven, The Netherlands.
Applications of rank-order-based methods to image and signal analyses have primarily focused on filtering. Classical median,
min, and max filters have long been part of standard image
processing toolboxes. More recent work has focused on more elaborate versions of such filters and associated computational issues. However, the application of these nonlinear methods to problems such as image interpretation has been scarce. We attempt to show that simple rank-order-based methods for coding image patches provide informative and computationally efficient local image descriptors.