Ptychographic coherent diffraction imaging is a novel lensless imaging method, which removes the resolution limitation by imaging elements in traditional optical lens imaging, so that its theoretical resolution is only limited by the X-ray wavelength and the numerical aperture of detector. However, the experiment background noise limits the further improvement of the imaging quality by this method, and may lead to the failure of image reconstruction. A new phase-retrieval iterative algorithm is proposed in this paper after studying the existing ptychographic iterative algorithms. The new algorithm uses the high redundancy of the ptychographic data set, and uses the gradient descent minimization method to reconstruct the background noise in synchronization with the reconstruction of the object and the detection, and realizes the blind separation of the noise and the signal. This algorithm is compared with the traditional iterative algorithms in simulation and experiment data reconstructions, demonstrating that this new algorithm can achieve a better signal-noise separation and a higher imaging quality.