A computationally efficient implementation of 2-D IQML

The 2-D IQML is a method for estimating the frequencies of 2-D sinusoids in noise. It is unique in the sense that it is the only 2-D method which exploits a linear representation of the orthogonal subspace. While this representation is structured and sparse, we have previously been unable to exploit this structure to devise a computationally efficient algorithm. This paper presents a remedy to this problem, yielding an efficient method for computing the 2-D likelihood function as expressed in terms of the orthogonal subspace. As such, the results impact not only the 2-D IQML but other estimators based on maximum likelihood as well.