崔玉敏, 尹丽菊, 隋立国, 周辉, 邓玉林. 光子计数集成成像的压缩与重构[J]. 应用光学, 2023, 44(2): 295-306. DOI: 10.5768/JAO202344.0202001
引用本文: 崔玉敏, 尹丽菊, 隋立国, 周辉, 邓玉林. 光子计数集成成像的压缩与重构[J]. 应用光学, 2023, 44(2): 295-306. DOI: 10.5768/JAO202344.0202001
CUI Yumin, YIN Liju, SUI Liguo, ZHOU Hui, DENG Yulin. Compression and reconstruction of photon counting integral imaging[J]. Journal of Applied Optics, 2023, 44(2): 295-306. DOI: 10.5768/JAO202344.0202001
Citation: CUI Yumin, YIN Liju, SUI Liguo, ZHOU Hui, DENG Yulin. Compression and reconstruction of photon counting integral imaging[J]. Journal of Applied Optics, 2023, 44(2): 295-306. DOI: 10.5768/JAO202344.0202001

光子计数集成成像的压缩与重构

Compression and reconstruction of photon counting integral imaging

  • 摘要: 集成成像技术作为一种重要的裸眼三维显示技术,在完整记录三维场景信息的同时,庞大的数据量给传输和存储带来了压力。为了实现图像的有效压缩和重构,根据光子计数集成成像的特点,基于分布式压缩感知理论,提出用于图像压缩与重构的方案。该方案将图像分为参考图像和非参考图像两类,对其设置不同的测量率并分别进行重构。为保证非参考图像的重构质量,提出一种联合重构算法。该算法首先对非参考图像进行分块测量,依据与参考图像之间的相关性进行图像块分类,然后结合参考图像测量值信息构建新的测量矢量,利用新的测量矢量完成初次图像重构。为了进一步提升图像重构质量,对初次重构结果进行二次残差补偿重构,获得最终重构结果。最后通过设置不同的测量率进行了大量实验,实验结果表明,所提算法在测量率为0.25时,图像重构质量可以达到30 dB,测量率为0.4时,图像质量可以达到35 dB,算法性能具有一定的优越性。

     

    Abstract: As an important naked eyes three-dimensional display technology, integral imaging technology completely records the three-dimensional scene information while the huge amount of data puts pressure on the transmission and storage. In order to achieve effective compression and reconstruction of images, according to the characteristics of photon counting integral imaging, a scheme for image compression and reconstruction based on the distributed compressed sensing theory was proposed. In this scheme, images were divided into reference images and non-reference images, which were set with different measurement rates and reconstructed respectively. To ensure the reconstruction quality of non-reference images, a joint reconstruction algorithm was proposed. Firstly, the block measurements were performed on the non-reference images, and the image blocks classification was carried out by considering the correlation with the reference images. Then, combined the measurement information of the reference images, a new measurement vector was constructed to complete the initial image reconstruction. In order to further improve the image reconstruction quality, the image was reconstructed with secondary residual compensation to obtain the final reconstruction results. Finally, a large number of experiments were conducted by setting different measurement rates. The experimental results show that the image reconstruction quality of the proposed algorithm can reach 30 dB when the measurement rate is 0.25, and can reach 35 dB when the measurement rate is 0.4. The performance of the algorithm has certain superiority.

     

/

返回文章
返回