High-speed computational multiphoton imaging through scattering media

To achieve high capacity, high speed, and secure image transmission, we propose a multi-image computational ghost imaging (CGI)-based encryption scheme that integrates compressed sensing (CS), block scrambling, and dynamic-salt-driven bidirectional XOR diffusion. First, multiple images are partitioned into 8 × 8 pixel blocks, and their spatial structure is disrupted through random scrambling. The scrambled composite image then undergoes pixel-level encryption via two-round bidirectional XOR diffusion, using session-unique keys derived from SHA-256-based dynamic salt, eliminating the statistical characteristics of the original images. Subsequently, each pixel block is subjected to both Gaussian CS and Hadamard-based CGI measurements in parallel, achieving dual-mode compressive encryption and enhancing robustness through measurement redundancy. Finally, only the scrambling key, the XOR-diffusion key, and the compressed measurements are stored; the original image information is thus transformed into unrecognizable measurement data. During the decryption process, the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) with a Discrete Cosine Transform (DCT) sparse basis is employed for dual-sparse reconstruction from the compressed measurements, recovering the encrypted composite image. An inverse XOR operation is then applied to remove the pixel-level diffusion, followed by block reordering using the scrambling key to restore the original images. Experimental results demonstrate that the proposed scheme enables efficient and secure multi-image transmission while maintaining high decrypted image quality. Security analysis indicates that the scheme possesses high key sensitivity, effectively resisting chosen-plaintext attacks. Histogram uniformity analysis and cropping attack resistance experiments further confirm its excellent statistical security and robustness.

International audience

The multi-frequency temporal phase unwrapping (MF-TPU) method, as a classical phase unwrapping algorithm for fringe projection techniques, has the ability to eliminate the phase ambiguities even while measuring spatially isolated scenes or the objects with discontinuous surfaces. For the simplest and most efficient case in MF-TPU, two groups of phase-shifting fringe patterns with different frequencies are used: the high-frequency one is applied for 3D reconstruction of the tested object and the unit-frequency one is used to assist phase unwrapping for the wrapped phase with high frequency. The final measurement precision or sensitivity is determined by the number of fringes used within the high-frequency pattern, under the precondition that its absolute phase can be successfully recovered without any fringe order errors. However, due to the non-negligible noises and other error sources in actual measurement, the frequency of the high-frequency fringes is generally restricted to about 16, resulting in limited measurement accuracy. On the other hand, using additional intermediate sets of fringe patterns can unwrap the phase with higher frequency, but at the expense of a prolonged pattern sequence. With recent developments and advancements of machine learning for computer vision and computational imaging, it can be demonstrated in this work that deep learning techniques can automatically realize TPU through supervised learning, as called deep learning-based temporal phase unwrapping (DL-TPU), which can substantially improve the unwrapping reliability compared with MF-TPU even under different types of error sources, e.g., intensity noise, low fringe modulation, projector nonlinearity, and motion artifacts. Furthermore, as far as we know, our method was demonstrated experimentally that the high-frequency phase with 64 periods can be directly and reliably unwrapped from one unit-frequency phase using DL-TPU. These results highlight that challenging issues in optical metrology can be potentially overcome through machine learning, opening new avenues to design powerful and extremely accurate high-speed 3D imaging systems ubiquitous in nowadays science, industry, and multimedia.

Imaging tools that allow for high resolution high-speed network-level brain activity mapping are of high interest in neuroscience and in other biomedical applications studying biological population dynamics in scattering tissues in vivo. Currently nonlinear optical microscopy plays a key role in imaging neural activity in behaving animals. In scattering tissue, two-photon microscopy in combination with raster scanning approaches have been widely applied, but one of the most difficult challenges is to achieve sufficient image contrast deep in scattering samples[1]. Due to limited number of emitted photons available for image reconstruction, accessing neuronal dynamics with sufficient temporal resolution for large-scale volumetric measurements of neural activity in scattering tissue is challenging. For high-speed raster scanning multiphoton imaging, fundamentally the speed limit is set by the pulse repetition rate of the femtosecond laser used, i.e. at least one laser pulse must be used per image pixel. In practice, this limit is also ruled by a compromise of the speed and inertia of the imaging and beam steering hardware, as well as the maximum allowed incident laser power on the sample to avoid damage or excessive heating. On the other hand, the available pulse peak intensity and operation wavelength of the driving laser source play also a significant role for achieving deep tissue imaging. To counteract the power decrease with depth, in order to maintain sufficient intensity at the focus at significant depths in scattering media, it is key to increase the energy of the excitation pulse. Yb-fiber chirped pulse amplifiers providing femtosecond, multi μJ-level pulses at very high repetition rates are very attractive for high-speed functional imaging. In this work, we present a μJ-level, MHz repetition rate all-fiber integrated chirped pulse amplifier delivering clean ∼180 fs pulses (Fig. 1 (a-c)), that enabled us to demonstrate a new approach for fast volumetric raster scanning microscopy, using temporal focusing [2] to produce an enlarged point spread function of 5×5×10 μm (matching the typical size of neurons in the mouse cortex) to allow for single laser pulse per vowel excitation [3]. This is important because it allows to sample the imaging volumes with the minimally required number of voxels, which leads to faster volume sampling and higher signal-to-noise ratios for the same average laser power. The large stretching ratio of the pulses in the Yb-fiber amplifier allows us to change the repetition rate and output energy of the system over a large range, without compromising the pulse fidelity. At MHz repetition rate, we have obtained up to 10 W of average power from the system using less than 60% of the available pump power, and at 100 kHz, compression of 1 W of output power (10 μ J pulse energy) from a similar system, using the same stretcher, but less efficient gratings yielded 160 fs pulses [4]. With the system set to a repetition rate of 4 MHz and 600 nJ output energy, we have performed high-speed volumetric functional imaging in the brain of living, awake mice, using a microscope with a sculpted focal volume of 5×5×10 μm3 (xyz) through temporal focusing. We demonstrate high-speed single neuron-resolution three-dimensional imaging in mouse brain in-vivo, on GCamp6 and jRGECKO1 labelled neurons, over large volumetric FOVs (up to 500×500×500 μm) at multi-Hertz (3–6 Hz) update rate, see Fig. 1 (d) and (e).

Single-pixel imaging uses a single-pixel detector, rather than a focal plane detector array, to image a scene. It provides advantages for applications such as multi-wavelength, three-dimensional imaging. However, low frame rates have been a major obstacle inhibiting the use of computational ghost imaging technique in wider applications since its invention one decade ago. To address this problem, a computational ghost imaging scheme, which utilizes an LED-based, high-speed illumination module is presented in this work. At 32 × 32 pixel resolution, the proof-of-principle system achieved continuous imaging with 1000 fps frame rate, approximately two orders larger than those of other existing ghost imaging systems. The proposed scheme provides a cost-effective and high-speed imaging technique for dynamic imaging applications.

This work discusses innovative high-speed imaging methods, merging hard-ware and algorithms in computational imaging. It focuses on techniques like rolling shutter, and coded aperture compressive temporal imaging for efficient data capture and high-quality compression.

1,000 fps motion vector (MV) estimation and classification engine for highspeed computational imaging in a 3D stacked imager/processor module is proposed, prototyped, assembled, and also tested. The module features 1) ThruChip interfaces for high fps image transfer, 2) orders of magnitude more area/power efficient MV estimation architecture compared to conventional ones, and 3) a cognitive classification scheme employed on MV patterns, enabling the classification of moving objects not possible in conventional proposals.

High-frame-rate two-photon imaging is becoming increasingly important in neuroscience for recording fast voltage and neurotransmitter signals from many sources at a time. I will summarize developments in this field and my lab’s recent efforts to build a DMD-based random-access two-photon microscope capable of recording thousands of sources at kilohertz framerates, without the need for computational image reconstruction.