A self-synchronized high speed computational ghost imaging system: A leap towards dynamic capturing

Abstract

a b s t r a c tHigh quality computational ghost imaging needs to acquire a large number of correlated measurements between the to-be-imaged scene and different reference patterns, thus ultra-high speed data acquisition is of crucial importance in real applications. To raise the acquisition efficiency, this paper reports a high speed computational ghost imaging system using a 20 kHz spatial light modulator together with a 2 MHz photodiode. Technically, the synchronization between such high frequency illumination and bucket detector needs nanosecond trigger precision, so the development of synchronization module is quite challenging. To handle this problem, we propose a simple and effective computational self-synchronization scheme by building a general mathematical model and introducing a high precision synchronization technique. The resulted efficiency is around 14 times faster than state-of-the-arts, and takes an important step towards ghost imaging of dynamic scenes. Besides, the proposed scheme is a general approach with high flexibility for readily incorporating other illuminators and detectors.

Anti-interference Computational Ghost Imaging with Pink Noise Speckle Patterns

“…The measurements are performed under several environmental interference situations. We also compare the results of pink noise with white noise which has been commonly used in the last decade [33].…”

Multispectral imaging using a single bucket detector

“…We utilize the novel self-synchronization technique in ref. 26 to synchronize the DMD and the detector. In all, it takes around 1 minute for data acquisition.…”

“…1 . Utilizing the fact that the response speed of a bucket detector (MHz or GHz) is magnitudes faster than that of a spatial light modulator (no higher than KHz) 15 23 26 , we encode the target scene’s spectral information into this speed gap. Specifically, the proposed MSPI technique introduces spectrum-dependent sinusoidal intensity modulation to illumination during the elapse of each spatially modulated pattern.…”

Efficient single-pixel multispectral imaging via non-mechanical spatio-spectral modulation

“…The SLM works at 10 kHz with 25% blank time (displaying an all-zero pattern) for synchronization, and the sampling rate of the single-pixel detector (Thorlabs PDA100A-EC Silicon photodiode, 340~1100 nm) is set to be 50 MHz. A self-synchronized sampling scheme 23 is utilized to synchronize the SLM and the detector. For each object, 4915 high-resolution patterns (equivalently, 39320 low-resolution Gaussian random binary patterns for 8 channels) are used for multispectral image reconstruction, and the corresponding compression ratio is 0.3 comparing to the full measurement of the spectral cube ( N = 128 × 128 × 8).…”