Xue-Feng Liu

Adaptive compressive ghost imaging based on wavelet trees and sparse representation.

Compressed sensing is a theory which can reconstruct an image almost perfectly with only a few measurements by finding its sparsest representation. However, the computation time consumed for large images may be a few hours or more. In this work, we both theoretically and experimentally demonstrate a method that combines the advantages of both adaptive computational ghost imaging and compressed sensing, which we call adaptive compressive ghost imaging, whereby both the reconstruction time and measurements required for any image size can be significantly reduced. The technique can be used to improve the performance of all computational ghost imaging protocols, especially when measuring ultra-weak or noisy signals, and can be extended to imaging applications at any wavelength.

Lensless ghost imaging with sunlight.

An experiment demonstrating lensless ghost imaging (GI) with sunlight has been performed. A narrow spectral line is first filtered out and its intensity correlation measured. With this true thermal light source, an object consisting of two holes is imaged. The realization of lensless GI with sunlight is a step forward toward the practical application of GI with ordinary daylight as the source of illumination.

Complementary compressive imaging for the telescopic system

Conventional single-pixel cameras recover images only from the data recorded in one arm of the digital micromirror device, with the light reflected to the other direction not to be collected. Actually, the sampling in these two reflection orientations is correlated with each other, in view of which we propose a sampling concept of complementary compressive imaging, for the first time to our knowledge. We use this method in a telescopic system and acquire images of a target at about 2.0 km range with 20 cm resolution, with the variance of the noise decreasing by half. The influence of the sampling rate and the integration time of photomultiplier tubes on the image quality is also investigated experimentally. It is evident that this technique has advantages of large field of view over a long distance, high-resolution, high imaging speed, high-quality imaging capabilities, and needs fewer measurements in total than any single-arm sampling, thus can be used to improve the performance of all compressive imaging schemes and opens up possibilities for new applications in the remote-sensing area.

Three-dimensional single-pixel compressive reflectivity imaging based on complementary modulation

A three-dimensional (3D) imager with a single-pixel detector and complementary intensity modulation of a digital micromirror device (DMD) array, which does not rely on scene raster scanning as in light detection and ranging (LIDAR) or on a two-dimensional array of sensors as used in time-of-flight (TOF) cameras, can not only capture full-color, high-quality images of real-life objects, but also recover the depth information and 3D reflectivity of the scene, reducing the required measurement dimension as well as the complexity, and cutting the cost of the detector array down to a single unit. The imager achieves spatial resolution using compressed sensing to exploit the sparsity of the signal. The disparity maps of the scene are reconstructed using sum of absolute or squared differences to reveal the depth information. This nonscanning, low-complexity 3D reflectivity imaging prototype may be of considerable value to various computer vision applications.

A double-threshold technique for fast time-correspondence imaging

We present a robust imaging method based on time-correspondence imaging and normalized ghost imaging (GI) that sets two thresholds to select the reference frame exposures for image reconstruction. This double-threshold time-correspondence imaging protocol always gives better quality and signal-to-noise ratio than previous GI schemes, and is insensitive to surrounding noise. Moreover, only simple add and minus operations are required while less data storage space and computing time are consumed; thus, faster imaging speeds are attainable. The protocol offers a general approach applicable to all GI techniques and marks a further step forward towards real-time practical applications of correlation imaging.

Protocol based on compressed sensing for high-speed authentication and cryptographic key distribution over a multiparty optical network

We present a protocol for the amplification and distribution of a one-time-pad cryptographic key over a point-to-multipoint optical network based on computational ghost imaging (GI) and compressed sensing (CS). It is shown experimentally that CS imaging can perform faster authentication and increase the key generation rate by an order of magnitude compared with the scheme using computational GI alone. The protocol is applicable for any number of legitimate user, thus, the scheme could be used in real intercity networks where high speed and high security are crucial.

Compressive microscopic imaging with “positive–negative” light modulation

An experiment demonstrating single-pixel single-arm complementary compressive microscopic ghost imaging based on a digital micromirror device (DMD) has been performed. To solve the difficulty of projecting speckles or modulated light patterns onto tiny biological objects, we instead focus the microscopic image onto the DMD. With this system, we have successfully obtained a magnified image of micron-sized objects illuminated by the microscope’s own incandescent lamp. The image quality of our scheme is more than an order of magnitude better than that obtained by conventional compressed sensing with the same total sampling rate, and moreover, the system is robust against intensity instabilities of the light source and may be used under very weak light conditions. Since only one reflection direction of the DMD is used, the other reflection arm is left open for future infrared light sampling. This represents a big step forward toward the practical application of compressive microscopic ghost imaging in the biological and material science fields.

Iterative denoising of ghost imaging

We present a new technique to denoise ghost imaging (GI) in which conventional intensity correlation GI and an iteration process have been combined to give an accurate estimate of the actual noise affecting image quality. The blurring influence of the speckle areas in the beam is reduced in the iteration by setting a threshold. It is shown that with an appropriate choice of threshold value, the quality of the iterative GI reconstructed image is much better than that of differential GI for the same number of measurements. This denoising method thus offers a very effective approach to promote the implementation of GI in real applications.

Edge detection based on gradient ghost imaging

We present an experimental demonstration of edge detection based on ghost imaging (GI) in the gradient domain. Through modification of a random light field, gradient GI (GGI) can directly give the edge of an object without needing the original image. As edges of real objects are usually sparser than the original objects, the signal-to-noise ratio (SNR) of the edge detection result will be dramatically enhanced, especially for large-area, high-transmittance objects. In this study, we experimentally perform one- and two-dimensional edge detection with a double-slit based on GI and GGI. The use of GGI improves the SNR significantly in both cases. Gray-scale objects are also studied by the use of simulation. The special advantages of GI will make the edge detection based on GGI be valuable in real applications.

High-visibility ghost imaging from artificially generated non-Gaussian intensity fluctuations

The image quality in ghost imaging is vital in practical applications. Through theoretical analysis, we find that for thermal light the average intensity as well as the fluctuations of an arbitrary incident field can greatly influence the image quality. Based on this, we suggest an easily realizable scheme to improve the visibility by generating speckles of non-Gaussian intensity distributions with a spatial light modulator. Numerical simulation demonstrates that this method can significantly improve the visibility, and the effect on the imaging resolution is also discussed. This method may thus be helpful in promoting the implementation of ghost imaging in real applications.