In 3D gradient echo (GRE) and echo planar imaging (EPI), strong macroscopic field gradients are observed at air/tissue interfaces. The respective field gradients lead to an apparent increase in intravoxel dephasing, and, subsequently, to signal loss or image distortion. We propose an analytical approximation and a consequent method to compute low and high resolution field maps over all field map regimes (small and large echo spacing). A number of approaches which compute field maps from reconstructed phase data rely upon optimized linear least square fit and complex division approaches owing to the simplicity of their implementation. Most of these techniques, however, have historically considered only the phase signal when computing off-resonance maps while ignoring magnitude data. This latter may be of notable interest since the presence of noise is well depicted and interpreted. The presence of noise and phase aliasing that increase with increasing echo time (TE) and echo spacing (ΔTE) may seriously challenge the off-resonance map accuracy. These techniques still remain subject to the trade-off during the choice of GRE sequences, TE and ΔTE. In this work, we explore a novel model that considers any type of TE and ΔTE regime (small or large) and high phase wraps complexity. The field offset is weighted by the magnitude signal decay quality, to make the field mapping procedure as noise independent as possible. The performance of the proposed method was tested using simulated, experimental phantoms and in vivo human studies. The proposed approach markedly outperforms conventional techniques. It provides a correction equivalent to that of the conventional techniques in regions with high SNR (20), yielding a mean error of about 0.1 Hz, but appearing more robust in regions with low SNR (10), such as near the sinus cavity and at the very edge of the brain (mean error less than 1 Hz), where phase wraps and noise are highly present. The proposed technique shows promise to enhance field map generation over any acquisition regime and in regions of both high and low SNR and it can be easily implemented for rapid computation and used in a clinical setting.