Two Dimensional Efficient Multiplier-Less Structures of Mobius Function for Ramanujan Filter Banks
Journal
IEEE TRANSACTIONS ON SIGNAL PROCESSING
Journal Volume
68
Pages
5079-5091
Date Issued
2020
Author(s)
Chang, Kuo-Wei
Abstract
Ramanujan's sum has been recently applied to detect different periodic components in 2D image. By filtering in frequency domain and obtaining the least common multiplier (lcm) of periodicity matrices, the basic 2D periodic pattern can be found. This paper introduces a multiplier-less structure to efficiently implement this process, based on the well-known number theoretic function called Möbius function. The relationship between impulse train and Ramanujan's sum is also clearly described, in both 1D and 2D cases.
SDGs
Type
journal article