Research On Multichannel Spectral Image Acquisition Method Of Painted Heritage

Spectral-based color characterization can achieve the device- and environment-independent acquisition of color information, which can reduce the color distortion caused by metamerism phenomenon in the traditional co lor ime trie-based color characterization. In other words, it indicates that spectral-based color characterization can be applied to characterize the real and objective color information in color recording, communication and reproduction. Based on this point of view, several key problems of the multichannel spectral image acquisition technology, such as filter selection and design, spectral reconstruction and training sample selection, were tackled in this work. According to the results, the existing technological process of the multichannel spectral image acquisition technology was optimized and a workflow involving the multispectral image acquisition of painted heritage was established. The workflow was tested by several typical color sample sets and the field experiment of the multispectral image acquisition of the murals in Mogao Grottoes in Dunhuang. Results indicated that the proposed workflow can achieve accurate multispectral image acquisition. The main work of this dissertation included the following several aspects.Based on the theory of color science, the color distortion resulting from the metamerism phenomenon in the traditional colorimetric-based color characterization was analyzed and the advantages of the spectral-based color characterization were discussed in detail. The current progress of the multispectral image acquisition technology was summarized;the architecture and the key technical difficulties of the technology were also discussed in this part.The color information acquisition process of the multispectral image acquisition system was investigated. The factors, such as dark current noise, camera response characteristic, illumination uniformity and pixel shift, that influenced the quality of the acquired image, were analyzed and the corresponding correction methods dealing with these factors were also discussed in this part. By studying the influence of lighting sources on the accuracy of multispectral image acquisition, the requirement of the light source was investigated in this work, which can provide a reference for selecting the appropriate light source.In order to solve the problem that the principal components have negative values, two color filter design methods based on the principal component analysis with the non-negative and the smoothing constraints respectively were proposed. The proposed methods were tested by the simulation experiment. Results indicate that the proposed two methods can achieve or even exceed the spectral reconstruction performance using the best filter combination obtained by the exhaustive method. Their performances were also robust to the case of the reconstruction target change.A method based on principal component analysis was proposed to deal with the problem of representative color sample selection in the multispectral image acquisition process. Taking the murals in Mogao Grottoes as the typical examples, a large number of color patches painted by mineral pigments were prepared to provide data support for high precision multispectral image acquisition. Then the representative color samples were selected with the proposed method. It showed that the proposed method was superior to the previous methods in both colorimetric and spectral reconstruction accuracy and can meet the requirement of high accurate spectral reconstruction.A spectral reconstruction algorithm based on the feature vector space was proposed to improve the accuracy of the multispectral image reconstruction. Herein the feature vector space was extracted by singular value decomposition and non negative matrix factorization. The proposed method was tested by reconstructing the in-situ captured mural images of the Mogao Grottoes in Dunhuang and a fairly good result was obtained. In addition, another spectral reconstruction algorithm based on the interim connection space (ICS) for spectral data was proposed. The method combines the excellent spectral reconstruction performance of ICS with the inverse distance weighted least square method, which can achieve the goal of high accuracy spectral reconstruction.