Abstract: This paper presents an AR model based maximum entropy method for magnetic resonance imaging.Since it has computedly effecient and better quality of reconstructed image,it will be an alternative method to conventional use of the fast Fourier Transform algorithm in the reconstruction of magnetic resonance images.It is acquired by computing the respective AR coefficients and linerar prediction error sequence.In practical uses,the application FFT leads to artifacts and lower resolution in the image due to the truncated spectral data sets,but our reconstruction algorithm is superior both in enhancing resolution and in removing truncation artifacts.