We propose to use multiphoton interferences of photons emitted from statistically independent thermal light sources in combination with linear optical detection techniques to reconstruct, i.e., image, arbitrary source geometries in one dimension with subclassical resolution. The scheme is an extension of earlier work [Phys. Rev. Lett. 109, 233603 (2012)] where N regularly spaced sources in one dimension were imaged by use of the N th-order intensity correlation function. Here, we generalize the scheme to reconstruct any number of independent thermal light sources at arbitrary separations in one dimension exploiting intensity correlation functions of order m ≥ 3. We present experimental results confirming the imaging protocol and provide a rigorous mathematical proof for the obtained subclassical resolution.Higher order interferences with photons emitted by statistically independent light sources are an active field of research with the potential to increase the resolution in spectroscopy, lithography and interferometry [1][2][3][4][5][6], as well as in imaging and microscopy [7][8][9][10][11][12][13][14][15][16][17]. So far, subclassical resolution has been achieved by using entangled photons [3,8], but it was also shown that initially uncorrelated light fields -non-classical as well as classical -can be employed for that purpose [13][14][15][16][17]. Recently, Oppel et al. presented a detection scheme that allows to determine the source distance d for an array of N equidistant thermal light sources (TLS) with subclassical resolution by measuring the N th-order spatial intensity correlation function [14].Here, we show that the scheme presented in [14] can be generalized to reconstruct, i.e., image, any number of independent TLS at arbitrary separations in one dimension by exploiting photon correlation functions of order m ≥ 3. Measuring higher order correlations enables to isolate the spatial frequencies of the setup allowing to determine the source distribution with a resolution below the classical Abbe limit. We outline the imaging protocol and present experimental results verifying the theoretical predictions. A physical explanation and rigorous mathematical proof of the protocol and the spatial frequency filtering process is given in the Supplemental Material.We assume N TLS aligned on a grid in one dimension with lattice constant d at arbitrary separations, such that |R l+1 − R l | = x l · d, with x l ∈ N, l = 1, . . . , N − 1. The source geometry is thus determined by the lattice constant d and the N − 1 adjacent source distances x = (x 1 , x 2 , . . . , x N −1 ), whereas the spatial frequencies of the system are given by the tuple of source pair distances {ξ} ≡ {(x 1 ); (x 1 + x 2 ); . . . ; (x l1 + · · · + x l2 ); . . . ; (x 1 + · · · + x N −1 )} (see Fig. 1).To access the set of spatial frequencies {ξ} we make use of the normalized spatial mth-order intensity correlation function g Here, : · : ρ denotes the (normally ordered) quantum mechanical expectation value for a system in the state ρ andÊ (−) (r j...