"We propose passive free-running SPAD (PF-SPAD) imaging, an imaging modality that uses SPADs for capturing 2D intensity images with unprecedented dynamic range under ambient lighting, without any active light source. Our key observation is that the precise inter-photon timing measured by a SPAD can be used for estimating scene brightness under ambient lighting conditions, even for very bright scenes. We develop a theoretical model for PF-SPAD imaging, and derive a scene brightness estimator based on the average time of darkness between successive photons detected by a PF-SPAD pixel. Our key insight is that due to the stochastic nature of photon arrivals, this estimator does not suffer from a hard saturation limit. Coupled with high sensitivity at low flux, this enables a PF-SPAD pixel to measure a wide range of scene brightness, from very low to very high, thereby achieving extreme dynamic range. We demonstrate an improvement of over 2 orders of magnitude over conventional sensors by imaging scenes spanning a dynamic range of 10^6:1."
Certainly interesting work. I am going to focus on the comparison to QIS. I appreciate QIS being included as a supplemental note 8 (!). But, I am not sure why the authors ignore both multibit mode and HDR mode of the QIS, as both have been published several times over the past few years. The QIS HDR mode is something one can also do with SPADs, as has been demonstrated, and gives very extended DR. Multibit operation (say 3-5b) is another way to achieve HDR, especially in HDR mode. It is too bad the authors don't compare against these modes. Perhaps that is a shortcoming of not doing peer review and instead rushing to publish.
ReplyDeleteThen there is the other part that has to do with things people also care about, like sensor resolution, power and cost. This should be included if you are going to compare different methods of image capture.
So while I cannot comment on the mathematical model at this time (and may never do so) I remain highly confident that the QIS will occupy a very comfortable niche in photon-counting imaging, including HDR.
Thank you for your comments. We are glad you found this paper interesting!
DeleteThe goal of this paper is to present the idea of using a SPAD in the free-running mode for passive flux sensing. We show that operating a SPAD in this mode has unique noise and saturation properties, and can lead to high dynamic range imaging.
Current SPAD array technology is still in a very early stage. It cannot compete with existing megapixel imaging sensors. Due to various practical constraints, the ideas proposed in our paper are not ready for "prime-time" yet. We hope that our paper will pique readers' interest in this new image sensing technique and promote further development of SPAD hardware technology.
Comparison against other types of HDR imaging sensors and methods was not our focus in this paper. We included some discussion on conventional (CCD/CMOS) sensor and QIS sensor only to provide intuition about how a PF-SPAD has unique and interesting characteristics (like the nonlinear response and adaptive photon rejection). Ultimately, we believe that each sensing technology is interesting in its own right, and will likely find its own niche where it provides the best performance. In the future, it will be interesting to compare PF-SPADs with QIS and other HDR sensors (not just theoretically, but also accounting for practical issues of cost and power dissipation, as you suggested).
We're grateful for feedback as we finalize this paper. We will present a peer-reviewed version at CVPR in LA in June 2019. We hope to meet you and your colleagues from Gigajot there!
Hi Atul, are the references correctly placed? Some listed references are missing in the text description.
ReplyDeleteHi, thank you for the feedback. I could not find a listed reference that isn't cited in the main text. Is there a specific number you couldn't find?
DeleteHowever, I see that the references are cited in random order which can be inconvenient to track down. When we post a revision (in a few weeks) I will do a more thorough check and also fix the citations so that they appear in sorted order in the text. (This was my first time posting LaTeX source to arXiv, so I'm still learning.)