The spectral camera based on ghost imaging via sparsity constraints (GISC spectral camera) is a phase modulated compressive snapshot spectral imager. It makes use of the second-order intensity correlation of the optical field to resolve the spatial and spectral information. In this paper, an optical design for GISC spectral camera which aims to obtain the desired spatial and spectral resolution is presented. A system calibration strategy based on few-mode optical fiber and monochrometer is developed. The snapshot spectral imaging experiments for the test targets and natural scenes are conducted using the prototype of GISC spectral camera loaded on the tethered balloon. The result of the spatial resolution, linearity and spectra reconstruction error of the prototype is quantitatively evaluated. The distinguishable size at the distance of 1km is around 0.34m. The linearity is higher than 0.99 among the wavelength channels from 410nm to 640nm. The reconstructed spectra of eight color targets are compared with those measured by a commercial spectroradiometer. The average relative root mean squared error (RRMSE) of the reconstructed spectra is 0.65.

Multispectral image reconstruction, which aims to recover a three-dimensional (3D) spatial–spectral signal from a two-dimensional measurement in a spectral camera based on ghost imaging via sparsity constraint (GISC), has been attracting much attention recently. However, faced with abundant 3D spectral data, the reconstruction quality cannot meet the visual requirements. Based on the robust data processing capability of deep learning, a novel network called SSTU-Net3+ is constructed by improving U-Net3+ with a spatial–spectral transformer (SST). To enhance the feature representation of images during reconstruction, mixed pooling modules and new convolution processes are proposed to improve the performance of the encoder and decoder, with U-Net3+ as the backbone. To boost the quality of reconstructed images, with split and concatenate (Concat) operations, we construct SST modules by exploiting both spatial and spectral correlations of multispectral images to refine the spatial and spectral features. Furthermore, we employ the SST in the decoder to reconstruct the desired 3D cube. Given similar network parameters, experiments on GISC spectral imaging data show that, compared to convolutional neural network–based methods, the average peak signal-to-noise ratio of images reconstructed using SSTU-Net3+ is improved by 3%, the structural similarity is enhanced by 3%, and the spectral angle mapping is cut by 12%. Particularly, compared to differential ghost imaging and compressed sensing, the reconstruction quality of SSTU-Net3+ has been significantly improved. SSTU-Net3+ can process a large amount of 3D multispectral image data more efficiently and construct the target image more accurately than the abovementioned methods.

To solve the problem of poor quality in ghost imaging via sparsity constraints (GISC) multispectral image reconstruction with correlation operations and compressed sensing algorithms under low sampling rate detection conditions, we propose an end-to-end deep-learning-based method. Based on the U-Net, Res2Net-SE-Conv is employed instead of convolutional blocks to extract local and global image features at a more fine-grained level while adaptively adjusting the channel feature response. The two-dimensional contextual transformer is constructed to fully use contextual correlation information to enhance the effectiveness of feature representations. We employ the two-dimensional contextual transformer in the decoder part, dubbed CoT-Unet, to reconstruct the desired 3D cube. The results show that compared with U-Net, TSA-Net based on spatial-spectral self-attention, the PSNR of reconstructed images by the CoT-Unet is improved by 5 dB and 3 dB, respectively, SSIM is improved by 0.23 and 0.07, and SAM is decreased by 0.06 and 0.58. Compared with conventional algorithms such as DGI and CS, our method significantly improves the quality of reconstructed images. Furthermore, the comparison results at 10%, 20%, and 30% sampling rates show that our approach has the best quality in reconstructing GISC multispectral images at low sampling rates.

A far field super-Rayleigh speckle modulation scheme for GISC spectral camera is proposed. The design and fabrication of the phase plate generating super Rayleigh speckles are carried out. Experimentally the contrast is basically greater than 1 from 550nm to 800nm, and reaches the maximum of 1.8 at 600nm.

We propose a hyperspectral ghost imaging camera based on a flat-field grating, which can improve spectral resolution of spectral camera and provide a possibility of optimizing measurement matrix according to light fields with different wavelengths.

HBT telescope based on self-correlation in spatial domain can realize lensless imaging in a single measurement, which overcomes the measurement limitations of HBT interferometry. The simulation and discussion verify its correctness and feasibility.