Abstract Hadamard coding spectral imaging technology is a computational spectral imaging technology that modulates the target’s spectral information and recovers the original spectrum by the inverse transformation. Compared with the dispersive spectrometer, this system has the advantage of better signal-to-noise ratio coming from multi-channel detection under low illumination. However, the coding process of this system is inevitability affected by several errors, including the misalignment of the coding template and the detector, scanning error, bad pixels, and so on. These errors would have an impact on the accuracy of the calculated spectrum. In this paper, we propose a unitive spectral reconstruction model under different errors and design an integrated approach to correct the above-mentioned errors simultaneously, including the bad pixel’s correction method with window function smoothing, the coding matrix’s correction method by using corrected template matrix to reconstruct coding matrix, and the push-scanning offset’s correction method including the inversion of line offset correction and column offset compensation, which could achieve better performance with the increase of spatial dimension. Experimental results on synthesized data and prototype tests show that the proposed correction method is effective in both single noise case and multiple noises condition, it is more accurate than traditional corrections in which only data preprocessing is finished.

中文翻译：

---

Hadamard 编码光谱成像仪光谱恢复的校正

摘要 阿达玛编码光谱成像技术是一种对目标光谱信息进行调制，通过逆变换恢复原始光谱的计算光谱成像技术。与色散光谱仪相比，该系统具有在低照度下多通道检测具有更好信噪比的优点。然而，该系统的编码过程不可避免地受到多种错误的影响，包括编码模板与检测器的错位、扫描错误、坏像素等。这些误差会影响计算光谱的准确性。在本文中，我们提出了不同误差下的统一谱重建模型，并设计了一种综合方法来同时校正上述误差，包括带窗函数平滑的坏像素校正方法，利用校正模板矩阵重构编码矩阵的编码矩阵校正方法，以及行偏移校正和列偏移补偿反相的推扫偏移校正方法，可以达到更好的效果。性能随着空间维度的增加而增加。合成数据的实验结果和原型测试表明，所提出的校正方法在单噪声和多噪声情况下均有效，比传统的仅完成数据预处理的校正更准确。推扫偏移的校正方法包括行偏移校正和列偏移补偿的反演，随着空间维数的增加可以获得更好的性能。合成数据的实验结果和原型测试表明，所提出的校正方法在单噪声和多噪声情况下均有效，比传统的仅完成数据预处理的校正更准确。推扫偏移的校正方法包括行偏移校正和列偏移补偿的反演，随着空间维数的增加可以获得更好的性能。合成数据的实验结果和原型测试表明，所提出的校正方法在单噪声和多噪声情况下均有效，比传统的仅完成数据预处理的校正更准确。