Abstract：Image phase retrieval is the problem of recovering the original image phase from a Short-time Fourier Transform(STFT)measurement without any phase information.Regarding the problem of phase information loss of an image during the practical image observation and transmission,an algorithm based on the image restoring uniqueness of short-time Fourier transform magnitude measurement is proposed.In this algorithm,an iterative initial value of the gradient descent method obtained by the improved least squares(LS)method is minimizing a non-convex loss function.Then the algorithm could solve the problem of nonconvex optimization of phase retrieval on image.In the experiment,under the interference of different noise types and intensity,the contour of the recovered image is still clear.Experimental results show that the algorithm can recover the original phase from STFT measurement and strong robustness.
2019, Vol. 40 
