Phase retrieval at all defocus distances

Phase retrieval is to calculate the phase from the direct information of the intensity of focus and defocus. Classical methods include the phase retrieval algorithm based on the transport of intensity equation (TIE) and iteration. The phase retrieval algorithm based on TIE can directly obtain the absolute phase without reference beam and phase unwrapping but is sensitive to noise. On the other hand, the intensity difference method needs to be used to approximate the intensity differential when solving the algorithm. Therefore, the defocus distance between intensity images has great influence on the accuracy of retrieval results. The phase retrieval algorithm based on TIE will be limited if it is extended to the phase retrieval of large objects. In order to widen its application range, an adaptive phase retrieval algorithm based on convolutional neural network under all defocus distances is proposed. The new algorithm consists of two important components: phase retrieval algorithm module and convolutional neural network optimization module. Firstly, the preliminary retrieval results are obtained by the phase retrieval algorithm module at different defocus distances, and then the accuracy of the results is further improved by the convolutional neural network module. The experimental results illustrate that the proposed algorithm is not only suitable for different defocus distances but also has good accuracy and stability.

Automated summary

Traditional phase retrieval algorithms based on TIE are sensitive to noise and rely on intensity difference methods to approximate intensity differentials, which are influenced by defocus distances. To address these issues, an adaptive phase retrieval algorithm based on convolutional neural networks is proposed, achieving high accuracy and stability in retrieving results under various defocus distances.