The quality of MR image is an important factor to control the accuracy of the diagnosis and treatment, so the image quality improvement methods, which can remove the noise and do not deteriorate the spatial resolution of images, are demanded in medical imaging field. There is an MR image reconstruction method, which applies the Fourier transform after a quadratic phase modulation in the Fresnel transform imaging technique. This method has a great property that can diffuse the noise contained in NMR signal by the quadratic phase modulation. If we prepare suitable area for diffusion of the noise by using zero-filling technique before reconstructing an image with this method, the spectrum of noise will evenly distribute over the zero filled area, and on the other hand, the image spectrum appear as an aspect of multi-resolution type image which locally distribute. After the noise was removed by a threshold filter in this aspect space, we return the data to the beginning signal space by doing a series of inverse processes, and reconstruct the denoised signal by using usual Fresnel transform image reconstruction technique. An MR image, which the noise is greatly improved, whereas the deterioration in spatial resolution is hardly caused, can be obtained. In this work, we present a new MR image quality improvement method that uses the property of noise diffusing in the Fresnel transform imaging technique; and we describe the effectiveness from simulations that are evaluated on SNR improvement and extent of deterioration in spatial resolution, and compare them with standard Wiener filtering and Wavelet Wiener filtering. Finally, we verify the powerfulness of the proposed method and it is applicable to phase-scrambled Fourier transform imaging technique through experiment.