Abstract
Computational ghost imaging systems reconstruct images using a single element detector, which measures the level of correlation between the scene and a set of projected patterns. The sequential nature of these measurements means that increasing the system frame-rate reduces the signal-to-noise ratio (SNR) of the captured images. Furthermore, a higher spatial resolution requires the projection of more patterns, and so both frame-rate and SNR suffer from the increase of the spatial resolution. In this work, we combat these limitations by developing a hybrid few-pixel imaging system that combines structured illumination with a quadrant photodiode detector. To further boost the SNR of our system, we employ digital micro-scanning of the projected patterns. Experimental results show that our proposed imaging system is capable of reconstructing images 4 times faster and with ~33% higher SNR than a conventional single-element computational ghost imaging system utilizing orthogonal Hadamard pattern projection. Our work demonstrates a computational imaging system in which there is a flexible trade-off between frame-rate, SNR and spatial resolution, and this trade-off can be optimized to match the requirements of different applications.
Highlights
Ghost imaging[1,2,3,4,5], a technique closely related to single-pixel imaging[6,7], is an alternative to conventional digital cameras based on a focal plane detector array
Experimental results show that the proposed imaging system reconstructed images 4 times faster and with an ~33% higher signal-to-noise ratio (SNR) than a conventional single-pixel computational ghost imaging system relying on orthogonal Hadamard patterns
The SNRs are calculated from the reconstructed images, to which no noise reduction filter is applied, i.e., there is no image post process other than constrained matrix inversion
Summary
Ghost imaging[1,2,3,4,5], a technique closely related to single-pixel imaging[6,7], is an alternative to conventional digital cameras based on a focal plane detector array. Ghost imaging offers advantages in a growing range of non-conventional applications such as wide spectrum imaging[8,9], depth mapping[10,11] and imaging with spatially variant and reconfigurable resolution[12,13,14] Despite these niche applications, the relatively low frame-rate and signal-to-noise ratio (SNR) of computational ghost imaging compared to imaging based on detector arrays has prevented its use from becoming more widespread. Attempts to increase the frame-rate of computational ghost imaging systems have generally focused on two strategies: (i) shortening the signal acquisition time by using fast spatial light modulators, such as digital micro-mirror devices (DMD)[8], LED arrays[21] or optical phase array[22] or (ii) reducing the total number of correlation measurements required to reconstruct an image by utilizing orthogonal sampling strategies[23,24] or compressive sensing[6,25], i.e. under-sampling a scene and using prior knowledge of the scene such as sparsity constraints to guide the image reconstruction. Experimental results show that the proposed imaging system reconstructed images 4 times faster and with an ~33% higher SNR than a conventional single-pixel computational ghost imaging system relying on orthogonal Hadamard patterns
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
