Magnetic particle imaging (MPI) is one of the medical imaging technologies that are being used for early diagnosis of cancer and cardiovascular diseases. In MPI, the magnetic field gradient and alternative magnetic fields are applied from outside the body, and then the magnetization signal is generated from magnetic nanoparticles (MNPs). Furthermore, an image reconstruction is performed using this magnetization signal. However, the image quality is degraded by interference signals generated from outside of the objective region, when the magnetic field is insufficient to saturate the magnetization of MNPs. In this study, the particle distribution is estimated iteratively from the error between the detected magnetization signal and the calculated signal that estimated the previous reconstructed image. In order to confirm the validity of this proposed method, we performed numerical experiments on the condition that complex particle distribution. As results, image quality was upgraded and the image resolution was approximately 1.0 mm in numerical experiment.