Correlative interferometry has been proposed for terahertz imaging in applications such as standoff detection of concealed explosives. It offers the advantage of low cost image reconstruction in the pupil plane given the current unavailability of ready-to-use focal plane imaging technology at these wavelengths. Current interferometric approaches are based on finding the inverse Fourier transform of the spatial correlation computed from field measurements at a distance. This paper proposes an image reconstruction approach that provides a constrained least squares fit between computed autocorrelation from sensor measurements and the expression for the far field autocorrelation for extended objects and line arrays.