Color filter design method for multi-channel spectral acquisition system

The spectral reconstruction precision of multi-channel spectral acquisition system based on the commercial digital camera is greatly dependent on the filters used to acquire the image. In order to obtain optimal filter pair, which has excellent spectral reconstruction performance and is applicable to all image scenes, a color filter design method based on the principal component analysis is proposed in this work. Firstly, a number of widely-used color filters are collected and their transmittances are measured and then written in matrix form. Secondly, the proposed method is adopted to extract the first two principal components of the matrix. The normalized principal components are the transmittances of the desired synthetic filters. To compare the spectral reconstruction performance of the synthetic filters with the other filter pair obtained by a common used method, two virtual spectral acquisition systems were constructed with two pairs of filters. The performance of each system was evaluated in terms of chromatic difference and spectral root mean square error. The results indicate that the synthetic filters obtained by the proposed method have a better performance than the common used methods. In addition, the performance of the synthetic filters is stable when the target image scenes change.