用快速采样与智能重建实现每秒6.8帧的高清单像素成像。
High-resolution single-pixel imaging in real time with iterative or deep learning-based reconstruction enhancement
- 结合特定采样策略与稀疏性增强重建，提升图像质量。
- 压缩比0.41%下实现6.8帧/秒的实时高清成像。
- 适配中端显卡，适合动态场景实时成像应用。
我们提出一种压缩单像素成像（SPI）框架，可在不足一秒内完成高分辨率图像采集。该框架融合专用采样策略与定制重建方法，实现对空间稀疏场景的原生1024x768分辨率高质量成像。重建过程分两阶段：首先利用测量矩阵的广义逆快速恢复图像；随后利用场景的空间稀疏性，在密集区域通过迭代方法或基于神经网络的方法进一步优化重建效果。在0.41%压缩比、DMD工作频率为22 kHz的条件下，系统支持6.8帧/秒的实时动态成像，重建速度匹配采集速率，可在中端桌面GPU上运行。
原文摘要 · Abstract (English)
We introduce a compressive single-pixel imaging (SPI) framework for high-resolution image capture in fractions of a second. This framework combines a dedicated sampling strategy with a tailored reconstruction method to enable high-quality imaging of spatially sparse scenes at the native 1024x768 resolution of a digital micromirror device (DMD). The reconstruction process consists of two phases: first, the measured data is processed using the generalized inverse of the measurement matrix for quick image recovery. Then, the spatial sparsity of the scene is leveraged to enhance reconstruction in dense areas, using either an iterative method or a neural network-based approach. With a compression ratio of 0.41% and an image acquisition rate of 6.8 Hz at 22 kHz DMD operation, this framework supports real-time, high-resolution dynamic imaging with the reconstruction that matches the acquisition rate on a mid-tier desktop GPU.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。