Abstract
Quantum ghost imaging offers many advantages over classical imaging, including the ability to probe an object with one wavelength and record the image with another (non-degenerate ghost imaging), but suffers from slow image reconstruction due to sparsity and probabilistic arrival positions of photons. Here, we propose a two-step deep learning approach to establish an optimal early stopping point based on object recognition, even for sparsely filled images. In step one we enhance the reconstructed image after every measurement by a deep convolutional auto-encoder, followed by step two in which a classifier is used to recognise the image. We test this approach on a non-degenerate ghost imaging setup while varying physical parameters such as the mask type and resolution. We achieved a fivefold decrease in image acquisition time at a recognition confidence of 75%. The significant reduction in experimental running time is an important step towards real-time ghost imaging, as well as object recognition with few photons, e.g., in the detection of light sensitive structures.
Highlights
Quantum ghost imaging offers many advantages over classical imaging, including the ability to probe an object with one wavelength and record the image with another, but suffers from slow image reconstruction due to sparsity and probabilistic arrival positions of photons
We know that numbers four and nine are two very written objects, prone to discrimination problems in machines, and we further tested the ability of the two-step approach to accurately discriminate between these digits
After imposing a strict evaluation criteria we determined that a recognition algorithm, individually, allows for the reduction of image reconstruction time to around 40% of the total reconstruction time
Summary
Our holograms (objects and masks) were created using MATLAB, for data acquisition we used LabView, all image processing techniques were carried out in JupyterLab using Python While we focused considerable efforts on adequate training and hyper-parameter tuning of the deep neural networks, we varied physical parameters within the experiment to test the robustness of our deep learning approach Two such parameters were the patterned mask set type and the mask resolution. After each measurement (patterned mask) we applied our two-step approach which consists of passing the reconstructed raw image through our auto-encoder for image enhancement by denoising, followed by passing the enhanced image into our neural classifier for object recognition. We have achieved image enhancement and recognition and can, safely say that the experiment can be stopped
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
