基于有限维模型的舌图像颜色校正方法可行性研究

在中医舌诊客观化的研究中，舌图像颜色数据的准确记录是后续计算机分析的基础。为消除光照环境和采集设备的变化对图像数据的影响，采集的图像必须进行颜色校正。在基于光谱反射率还原的颜色校正方法中，有限维模型是光谱反射率降维的有效方法。本文从光谱反射率复原角度，对舌色范围进行了有限维模型分析，并初步分析了有限维模型近似舌色光谱反射率的误差，结果表明前三个基函数的特征值能量高达99．13％，由其估计的光谱反射率曲线计算得到的颜色值与实际值比较接近，由此证实了基于有限维模型的舌图像颜色校正方法具有一定的可行性。

Finite Dimension Model Based Tongue Color Correction

In characterizing a tongue diagnosis, the veracity of tongue image color data is the basis for a correct image analysis. Images, therefore, need to be color corrected, to offset unneeded effects offered by different illumination and image capturing devices. In spectral reflectance reproduction based color correction, the finite dimension model can be employed to condense spectral reflectance. In the study, tongue color is analyzed using the finite dimension model. Authors have worked out a solution to correct the errors caused by approximation of the finite dimension model. Experiment results show an overall variance of 99. 13% generated by three basic functions, presenting a good approximation to the spectral reflectance of tongue. The finite dimension model based tongue color correction is proved feasible.