XZXin Zhou

Xin Zhou

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Xin Zhou has a total of 102 co-authors. Here are the top co-authors by number of shared publications:

  • Xing Bai

    Department of Opto-Electronics Science and Technology, Sichuan University, Chengdu, China

    16 shared publications
  • Sheng Yuan

    North China University of Water Resources and Electric Power

    15 shared publications
  • Yang Liu

    Department of Opto-Electronics Science and Technology, Sichuan University, No.24, Section 1, South Yihuan Road, Chengdu, 610065, CHINA

    11 shared publications
  • Mingze Sun

    Sichuan Univ, Dept Optoelect Sci & Technol, Chengdu 610065, Peoples R China

    10 shared publications
  • Xingyu Chen

    Department of Opto-Electronics Science and Technology, Sichuan University, Chengdu, 610065, People’s Republic of China

    10 shared publications
  • Yujie Wang

    Department of Opto-Electronics Science and Technology, Sichuan University, Chengdu, 610065, People’s Republic of China

    10 shared publications

Latest Publications

  • A low-storage-overhead approximate thumbnail-preserving encryption scheme based on grayscale sparse speckle single-pixel imaging

    by Daili Wang, Yi Wei, Yu Kou, Ying Li, Dongming Huo, Sheng Yuan, Xin Zhou

    Thumbnail-preserving encryption (TPE) provides a practical solution for privacy protection and visual usability of cloud images. However, most existing TPE schemes still suffer from a high file expansion rate of the encrypted image, a relatively complex encryption process, and limited efficiency. To this end, this paper treats the non-zero elements of the speckle in single-pixel imaging as weights, pre-sets the sum of the weights, and proposes a novel approximate TPE scheme. Under low-sampling conditions, the scheme constructs a grayscale sparse measurement matrix, treats the non-zero elements of the speckle as weights, and constrains the sum of the weights to 1, so that the obtained bucket signals are equivalent to the weighted mean of multiple pixels within a block. The encrypted image directly formed by the bucket signals preserves an approximate thumbnail of the original image without additional processing. Meanwhile, thanks to the flexible values of the non-zero elements in the grayscale sparse speckle, a huge key space can be provided even with a small block size. In addition, by exploiting the principle of compressive sensing, the proposed scheme effectively reduces the number of pixels in the encrypted image, significantly lowers the ciphertext expansion rate, and achieves lower cloud storage overhead. Simulation and experimental results demonstrate that the scheme achieves a millisecond-level encryption time, a ciphertext expansion rate below 0.5 at a 25% sampling rate, and a PSNR of 33 dB for the decrypted image when the block size is 64 × 64, indicating that the scheme still maintains good decryption performance under low-sampling conditions.

  • Task-driven single-pixel salient object detection via deep semantic compression

    by Ying Li, Yi Wei, Chao Yang, Zhenghua Hu, Yu Kou, Sheng Yuan, Xin Zhou

    Single-pixel imaging (SPI) is a cost-effective computational imaging technique in special spectral bands. Traditional SPI-based sensing generally follows an “imaging-first” strategy, where reconstructed images may contain redundant information, such as irrelevant backgrounds. Consequently, this redundancy incurs additional hardware and computational costs during data acquisition, transmission, and storage. To mitigate this waste of system resources, performing advanced sensing tasks directly from 1D measurements is crucial. Therefore, we propose a task-driven, end-to-end single-pixel salient object detection (SOD) method that achieves accurate saliency detection directly from a small number of measurements, without explicit and complex image reconstruction. Specifically, the framework models the physical sampling process of SPI through a learnable task-aware sampling encoder, which modulates the target scene to obtain a compressed measurement vector. Subsequently, this vector is mapped into a 2D spatial representation via a fully connected layer and fed into the back-end network for pixel-level prediction. By jointly optimizing the learnable sampling layer and the back-end SOD network, the sampling matrix evolves into specific patterns tailored for SOD. Both simulation and experimental results demonstrate that the proposed method achieves robust and accurate saliency detection even at a low sampling rate of 6.25%. This framework provides a task-oriented paradigm for single-pixel SOD in environments with limited resources.

  • Single-pixel imaging based on row-orthogonal matrices

    by Zongguang Ouyang, Yi Wei, Daili Wang, Yu Kou, Ying Li, Sheng Yuan, Xin Zhou

    Single-pixel imaging, as a computational imaging technique based on non-pixelated detection, exhibits unique advantages in low-light, broadband, and special-band imaging. However, its widespread application is often constrained by the trade-off between imaging efficiency and reconstruction quality. In this paper, we propose a row-orthogonal matrix construction method based on iterative Kronecker products. By selecting different types of submatrices, a variety of row-orthogonal matrices can be flexibly generated, providing a novel structured speckle generation scheme to address the aforementioned trade-off. Moreover, the projection order plays a critical role in reconstruction quality. To this end, this paper further introduces a sorting strategy based on Fourier band energy comparison to optimize the speckle projection sequence. Through comparative simulations of different submatrix types, the more optimal matrix structures were identified. Comprehensive simulations and physical experiments demonstrate that, compared with random speckles, Walsh speckles, and Hadamard speckles, the proposed method achieves superior and more stable reconstruction quality for images of varying complexity at resolutions of 64 × 64 and 128 × 128, with particularly pronounced advantages under low sampling rates. Noise robustness analysis further confirms that the proposed approach maintains the highest structural similarity in the presence of added Gaussian noise, indicating strong robustness. This study provides an effective structured speckle generation and optimization scheme for efficient and high-quality single-pixel imaging.

  • Approximate thumbnail-preserving encryption based on single-pixel imaging

    by Daili Wang, Dongming Huo, Yi Wei, Zongguang Ouyang, Ying Li, Yu Kou, Chao Yang, Zhenghua Hu, Sheng Yuan, Xin Zhou

    With the rapid development of smart devices and Internet technology, the volume of visual data stored in the cloud has increased dramatically, posing severe challenges to security and privacy protection. Traditional image encryption methods often produce encrypted images that are difficult to view, preventing users from previewing content in the cloud. Thumbnail-preserving encryption (TPE) enables encrypted images to retain visual features, thereby striking a balance between security and usability. Most thumbnail-preserving encryption schemes employ complex pixel-wise encryption operations, which often suffer from long encryption times, low perceptual quality of the encrypted images, and a high file expansion rate of the encrypted image. This paper proposes an approximate thumbnail-preserving encryption scheme based on single-pixel imaging, referred to as TPE-SPI. The scheme utilizes the bucket signals, speckle patterns, and the pixel values of the target image, along with the relationships among them, to achieve efficient and secure thumbnail-preserving encryption through simple processing of the bucket signals. Compared to other schemes, TPE-SPI offers higher encryption efficiency and better visual quality of the encrypted image, while effectively reducing the file expansion rate of the encrypted image.


  • Single-pixel panoramic imaging based on Walsh-Hadamard basis

    by Yi Wei, Zhan Yu, Ying Li, Yu Kou, Zongguang OuYang, Daili Wang, Sheng Yuan, Xin Zhou

    Single-pixel panoramic imaging (SPPI) technology, which integrates convex mirrors with computational imaging algorithms, achieves efficient panoramic perception and demonstrates application potential in autonomous driving perception, pipeline inspection and low-altitude related applications. However, its algorithmic efficiency and accuracy still require optimization. This study improves the SPPI system from two aspects: modulation mode and reconstruction algorithm. In terms of modulation, the Walsh-Hadamard (W-H) orthogonal basis structure pattern is introduced, which effectively alleviates the noise sensitivity problem of the random mode in SPPI. In terms of algorithms, this study adapts and optimizes two key approaches for panoramic scenes: the orthogonal algorithm based on two-dimensional (2D) W-H transformation and the 2D compressive sensing (2DCS) algorithm based on the 2D Variational Augmented Lagrange Multiplier (V2DALM). These approaches enable independent asymmetric sampling in the radial and angular dimensions, allowing flexible configuration of sampling resources according to the characteristics of different scenes. The experimental system adopts a passive imaging mode, further enhancing the practicality and adaptability of SPPI. This study provides a scalable theoretical support for single-pixel panoramic perception systems and corresponding SPPI schemes for three typical requirements: high-fidelity imaging, high-speed reconstruction, and speed-fidelity trade-off. Finally, the SPPI performance of the V2DALM algorithm under different modulation patterns has been preliminarily analyzed and discussed.


  • Fresnel-DRPE-Cloud: An Optical Encryption Framework for Tunable Privacy-Usability Balance in Secure Cloud Storage

    by Dongming Huo, Hanwen Wang, Chang Cheng, Guangxiang Ji, Xiaoqing Song, Lisheng Wei, Chao Han, Xin Zhou

    This paper proposes Fresnel-DRPE-Cloud, an optical encryption framework for cloud-based image storage that resolves the privacy-usability tradeoff through tunable security. By introducing phase modulation parameters (α1, α2) into Fresnel-domain double random phase encoding (DRPE), the scheme enables dynamic control of encrypted image recognizability: classical noise-like encryption at α1 = α2 = 1.0 transitions to visually recognizability encrypted image as α values decrease. Hyperchaotic phase masks initialized by SHA-256 image hashes ensure plaintext-dependent security, achieving NPCR = 99.25% and UACI = 21.35% against differential attacks. For consumer technologies, this approach directly addresses critical user-experience limitations in encrypted media storage: it enables secure thumbnail browsing on mobile devices and smart displays while eliminating the need for frequent decryption—a key requirement for resource-constrained consumer electronics. Holistic image processing eliminates block partitioning - a core requirement in digital thumbnail-preserving encryption (TPE) - reducing encryption/decryption time by 1-2 orders of magnitude speedup compared to conventional block-based approaches. Decryption achieves near-perfect reconstruction (PSNR = 307.21 dB, SSIM = 1.00). This work establishes a new paradigm for optical encryption in cloud environments.


  • Sorting retrieval method for image ciphertexts based on ghost imaging

    by Dan Zhao, Jialin Zhang, Yue Li, Ying Li, Yi Wei, Zongguang Ouyang, Daili Wang, Sheng Yuan, Xin Zhou

    In this paper, a sorting retrieval method for image ciphertexts based on ghost imaging is proposed. With this scheme, the ciphertexts can be sorted according to similarity with the base image without decrypting the ciphertexts, and the sorting results obtained are highly consistent with the result of the plaintexts under the same sorting method, it is potential for applications in the field of homomorphic encryption. Differently from the traditional method of retrieval through deep learning, neural network and other models with a lot of pre-training, the proposed method in this paper does not require such training, but directly calculates the similarity between the ciphertexts and sort them, avoiding the additional computational overhead, and eliminating the risk of information leakage during pre-training. The proposed scheme is verified by numerical simulation using three datasets, MNIST, MNIST-L, and MNIST-C, the ciphertext sorting retrieval average accuracies can reach up to 96.5%, 98%, and 98%, respectively, showing the efficiency of this sorted retrieval method for image ciphertexts encrypted based on ghost imaging.

  • Optical encoding and multiplexing based regional single-pixel imaging via TD-SCDMA

    by Ying Li, Yu Kou, Yue Li, Jialin Zhang, Dan Zhao, Sheng Yuan, Zhijian Zhang, Xin Zhou

    Different from traditional imaging methods that rely on plane array detectors, single-pixel imaging (SPI) utilizes a single detector without spatial resolution to acquire images. However, achieving high-resolution reconstruction with SPI typically requires a large number of measurements, which limits the development of SPI. To achieve a balance between high resolution and frame rate, an SPI method based on optical encoding and multiplexing is proposed and demonstrated. In this method, the beam modulated by illumination patterns is subjected to orthogonal space-time encoding to achieve the redistribution of sampling density across different sub-regions of the target. After optical multiplexing, the combined light intensity is collected by a single-pixel detector and decoded to reconstruct each sub-region. Consequently, sub-regions assigned higher sampling densities exhibit improved detail resolution, more closely matching human visual perception compared to those with lower sampling. This work overcomes the limitation of traditional SPI approaches that impose uniform resolution across the entire field of view. With the introduction of space-time encoding, sampling density can be flexibly allocated among various regions. This enables the reconstruction of large-scale scenes at a low overall sampling rate while still achieving higher resolution in selected regions of interest. Moreover, the proposed method can be integrated with various reconstruction algorithms to further reduce computational costs and acquisition time.

  • Open Access

    Mid-infrared data encryption and transmission beyond detector bandwidth based on frequency down-conversion temporal ghost imaging

    by Linzhen He, Bo Hu, Zunwang Bo, Chenlin Zhang, Xin Zhou, Shensheng Han, Houkun Liang, Han Wu

    Computational temporal ghost imaging (CTGI) allows for retrieving the fast temporal object using a slow photodetector and can naturally encrypt the transmission data to enhance communication security. Here, we first demonstrate free-space data encryption based on frequency down-conversion CTGI in the mid-infrared (MIR) atmospheric window, where commercial fast photodetectors are lacking. The target temporal signal is digitally encrypted by a secret key and modulated at 1.5 μm light by an acousto-optic intensity modulator (AOM). The encrypted signal is then optically transferred to MIR light through nonlinear frequency down-conversion. The data decryption can be realized from the correlation operation between the intensity recorded by a 1 MHz bandwidth MIR detector and the secret key. Moreover, when the data rate is significantly higher than the bandwidth of the AOM, the modulated optical signal will be severely distorted, which adds extra physical encryption. As a result, we experimentally demonstrate MIR data encryption and transmission over a 3 m free-space link at a speed more than 60 times higher than the MIR detector bandwidth in a 3.3–3.5 μm spectral region. We anticipate the frequency down-conversion CTGI could provide a new platform for secure free-space communication in the MIR region.