针对漫射模型在近光源区具有较大模型误差, 导致光学特性参量反演准确度较低的问题, 提出了一种利用矩变换, 改变原始数据形态, 提高反演准确度的方法.比较了不同阶次条件下的光学特性参量反演准确度和拟合残差的统计分布状况, 分析了不同信噪比下重构系数的相对误差.对25组数据的研究结果表明：3阶矩变换可显著提高两个光学参量(μa、μ′s)的估计准确度.在无噪音条件下, 吸收系数μa重构的平均相对误差为7.04%;有效散射系数μ′s重构的平均相对误差为5.55%, 相比于自然对数变换, μa降低了8.57%;μ′s降低了32.73%.在信号噪音满足一定条件情况下(大于50 dB), 3阶矩变换仍然能获得较高的光学参量反演准确度.矩变换方法能有效地提高光学特性参量反演估计准确度.

Diffusion model has a disadvantage that exist a large model error in the near light source which lead to a lower accuracy for inversion of optical properties parameters. Therefore, the article proposed a moment-transformation method, which change the form of raw data and improve inverse accuracy. The article compared inverse accuracy of the optical properties under different orders conditions and statistical distribution of fit residuals and analyzed relative error of the reconstruction coefficient under different signal to noise ratio. The results of all set data indicate: 3 order moment-transformation can significantly improve estimation accuracy both optical parameters that 7.04% for μa and 5.55% for μ′s under the no noise condition. three order moment-transformation compared to original and the natural logarithm data transformation, the relatively error decrease 57.59%、8.57% for μa and 42.55%、32.73% for μ′s, respectively. In the case of the certain signal to noise ratio (greater than 50 dB), three order moment-transformation still get high inversion accuracy of optical parameter. Therefore, based on the distance weighted data transformation is an effective means of measuring optical parameters.