On the application of light field reconstruction for statistical object recognition

In this paper we apply light field reconstruction and rendering of object views to the problem of automatic generation of training material for a statistical object recognition system. The advantages of using a light field instead of real images are shown. We evaluate with respect to the error rate of the classifier, whether the reconstructed light field can be applied to the training step. We also show how the recognition rate of the classifier trained by the light field depends on its resolution.