Heterogeneous camera array for multispectral light field imaging

Multispectral light field acquisition is challenging due to the increased dimensionality of the problem. In this paper, inspired by anaglyph theory (i.e. the ability of human eyes to synthesize colored stereo perception from color-complementary (such as red and cyan) views), we propose to capture the multispectral light field using multiple cameras with di erent wide band filters. A convolutional neural network is used to extract the joint information of di erent spectral channels and to pair the cross-channel images. In our experiment, results on both synthetic data and real data captured by our prototype system validate the e ectiveness and accuracy of proposed method.