Objective To develop a new set of algorithms for high-resolution cellular traction force recovery based on two-dimensional Fourier domain by addressing the ill-posed nature of classic cellular force traction recovery. Methods By exploring the inherent characteristics and rules of displacement data on the substrates and Green’s function in the Fourier domain, the phenomenon of ill-posed deconvolution arising in cellular force traction recovery was investigated and a set of self-adaptive filtering algorithms was consequently developed to remarkably restrain the high frequency noise amplification. Results The ill-posed nature of classical Fourier transform traction cytometry (FTTC) made cellular traction force recovery extremely unstable, especially for relatively dense displacement data sampling. In contrast, the proposed self-adaptive filtering algorithms based on FTTC could make cellular traction force distribution more stable and reliable, as the effect of high frequency noise in displacement field on recovery results was weakened significantly. Conclusions This new technique for cellular traction force recovery can effectively suppress the noise and therefore improve the stability of force recovery procedure and spatial resolution, which is expected to find wider application in the study of cell substrate interactions.