Low Photon Count Phase Retrieval Using Deep Learning
Imaging systems’ performance at low light intensity is affected by shot noise, which becomes increasingly strong as the power of the light source decreases. In this Letter, we experimentally demonstrate the use of deep neural networks to recover objects illuminated with weak light and demonstrate be...
| Main Authors: | , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
American Physical Society
2019
|
| Online Access: | http://hdl.handle.net/1721.1/119896 https://orcid.org/0000-0002-7836-0431 https://orcid.org/0000-0002-4140-1404 |
| _version_ | 1826200583502036992 |
|---|---|
| author | Goy, Alexandre Sydney Robert Arthur, Kwabena K. Li, Shuai Barbastathis, George |
| author2 | Massachusetts Institute of Technology. Department of Mechanical Engineering |
| author_facet | Massachusetts Institute of Technology. Department of Mechanical Engineering Goy, Alexandre Sydney Robert Arthur, Kwabena K. Li, Shuai Barbastathis, George |
| author_sort | Goy, Alexandre Sydney Robert |
| collection | MIT |
| description | Imaging systems’ performance at low light intensity is affected by shot noise, which becomes increasingly strong as the power of the light source decreases. In this Letter, we experimentally demonstrate the use of deep neural networks to recover objects illuminated with weak light and demonstrate better performance than with the classical Gerchberg-Saxton phase retrieval algorithm for equivalent signal over noise ratio. The prior contained in the training image set can be leveraged by the deep neural network to detect features with a signal over noise ratio close to one. We apply this principle to a phase retrieval problem and show successful recovery of the object’s most salient features with as little as one photon per detector pixel on average in the illumination beam. We also show that the phase reconstruction is significantly improved by training the neural network with an initial estimate of the object, as opposed to training it with the raw intensity measurement. |
| first_indexed | 2024-09-23T11:38:38Z |
| format | Article |
| id | mit-1721.1/119896 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T11:38:38Z |
| publishDate | 2019 |
| publisher | American Physical Society |
| record_format | dspace |
| spelling | mit-1721.1/1198962022-10-01T05:00:27Z Low Photon Count Phase Retrieval Using Deep Learning Goy, Alexandre Sydney Robert Arthur, Kwabena K. Li, Shuai Barbastathis, George Massachusetts Institute of Technology. Department of Mechanical Engineering Goy, Alexandre Sydney Robert Arthur, Kwabena K. Li, Shuai Barbastathis, George Imaging systems’ performance at low light intensity is affected by shot noise, which becomes increasingly strong as the power of the light source decreases. In this Letter, we experimentally demonstrate the use of deep neural networks to recover objects illuminated with weak light and demonstrate better performance than with the classical Gerchberg-Saxton phase retrieval algorithm for equivalent signal over noise ratio. The prior contained in the training image set can be leveraged by the deep neural network to detect features with a signal over noise ratio close to one. We apply this principle to a phase retrieval problem and show successful recovery of the object’s most salient features with as little as one photon per detector pixel on average in the illumination beam. We also show that the phase reconstruction is significantly improved by training the neural network with an initial estimate of the object, as opposed to training it with the raw intensity measurement. 2019-01-09T19:54:57Z 2019-01-09T19:54:57Z 2018-12 2018-06 2018-12-12T18:00:20Z Article http://purl.org/eprint/type/JournalArticle 0031-9007 1079-7114 http://hdl.handle.net/1721.1/119896 Goy, Alexandre, et al. “Low Photon Count Phase Retrieval Using Deep Learning.” Physical Review Letters, vol. 121, no. 24, Dec. 2018. © 2018 American Physical Society https://orcid.org/0000-0002-7836-0431 https://orcid.org/0000-0002-4140-1404 en http://dx.doi.org/10.1103/PhysRevLett.121.243902 Physical Review Letters Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. American Physical Society application/pdf American Physical Society American Physical Society |
| spellingShingle | Goy, Alexandre Sydney Robert Arthur, Kwabena K. Li, Shuai Barbastathis, George Low Photon Count Phase Retrieval Using Deep Learning |
| title | Low Photon Count Phase Retrieval Using Deep Learning |
| title_full | Low Photon Count Phase Retrieval Using Deep Learning |
| title_fullStr | Low Photon Count Phase Retrieval Using Deep Learning |
| title_full_unstemmed | Low Photon Count Phase Retrieval Using Deep Learning |
| title_short | Low Photon Count Phase Retrieval Using Deep Learning |
| title_sort | low photon count phase retrieval using deep learning |
| url | http://hdl.handle.net/1721.1/119896 https://orcid.org/0000-0002-7836-0431 https://orcid.org/0000-0002-4140-1404 |
| work_keys_str_mv | AT goyalexandresydneyrobert lowphotoncountphaseretrievalusingdeeplearning AT arthurkwabenak lowphotoncountphaseretrievalusingdeeplearning AT lishuai lowphotoncountphaseretrievalusingdeeplearning AT barbastathisgeorge lowphotoncountphaseretrievalusingdeeplearning |