Recovery of spectral reflectances of the imaged objects without prior knowledge and color reproduction

艺术绘画和物品的数字化档案是将其记录下来并作为历史遗产留给未来的关键技术。由于绘画的印象随着光照的变化而发生极大的变化，因此考虑到视觉系统的色彩适应性，在各种光照下再现彩色图像是非常重要的。因此，利用图像数据准确地恢复成像物体的光谱反射率是非常重要的。本项目建立了一个模型，利用学习颜色样本及其对应的图像数据的光谱反射率来估计图像采集系统的噪声方差，并将其应用于在不事先知道成像物体光谱反射率的情况下，通过维纳估计来恢复光谱反射率。通过与其他恢复模型的光谱反射率恢复精度的比较，证实了在测试样本与学习样本不同的情况下，该方法的结果最准确。我们还建立了一个模型来评估一组旨在获取比色信息和光谱信息的传感器，并应用该模型来开发多光谱相机的最佳七通道光谱灵敏度。最后，利用多光谱相机对多幅艺术画作的光谱反射率进行恢复，并在不同照明条件下再现彩色图像

Digital archive of art paintings and objects is a key technology to record and to leave them as historical heritages for the future. Since the impressions of a painting change extremely by the changes in the illuminations, the reproduction of a color image under a variety of illuminations by taking account of the chromatic adaptation of the visual system is very important. Therefore, the accurate recovery of spectral reflectances of the imaged objects through the use of image data is very important for the purposeIn this project, we developed a model to estimate the noise variance of an image acquisition system by using the spectral reflectances of the learning color samples and their corresponding image data, and applied it to recover spectral reflectances by the Wiener estimation without prior knowledge of the spectral reflectances of the imaged objects. The accuracy of the recovered spectral reflectances by this proposal was compared with other recovery models and it is confirmed that the results by this proposal is the most accurate when test samples are different from learning samples. We also developed a model for the evaluation of a set of sensors aimed at the acquisition of colorimetric information and spectral information, and applied the model to develop the optimal seven channel spectral sensitivities of a multispectral camera.Finally, the spectral reflectances of the several art paintings were recovered by using the multispectral camera, and then color images were reproduced under a variety of illuminations