Performance of a MVE algorithm for compound eye image reconstruction using lens diversity

Reconstruction algorithms to compute a single improved resolution image from multiple lower resolution images have application in the design of cameras with flat form factors. The accuracy of these reconstructions will depend on measurement noise, measurement quantization, the structure of the image acquisition system, and the accuracy of the image acquisition model. This paper compares the expected and simulated performance for reconstructions from multiple lower resolution images. The analysis shows that designs using lenses with different imaging characteristics significantly improve the theoretical performance results. In addition, lens diversity allows the reconstruction problem to be naturally partitioned into a set of loosely coupled smaller reconstructions that are computationally more manageable.