A study on the reduction of reconstruction errors in CGH by simulated annealing

Abstract

In this thesis, a novel simulated annealing algorithm that can be used for the hologram ID pattern generation and greatly reduce reconstruction error rate compare to conventional simulated annealing algorithm is newly proposed. Low storage capacity and low data security have been main limitations of One-dimensional and two-dimensional barcode technology that are most widely used as identification technology. In this thesis, the concept of hologram ID that can overcome these shortcomings of previous barcode technologies has been proposed and the process of generation and recognition of hologram ID pattern has also been presented. Conventional simulated annealing algorithm has been adopted and analyzed for the purpose of hologram ID pattern generation. But the reconstruction error rate saturated around 10%. So the analysis of conventional simulated annealing algorithm has been performed to find out the reason of saturation of the reconstruction error rate. And we could recognize that there were much waste of computational power in conventional simulated annealing method while optimizing kinoform and this phenomenon has induced the saturation of reconstruction ER. Based on the result of previous analysis, a new simulated annealing algorithm has been proposed that can greatly reduce reconstruction error rate down to 0.04%. The results, such as saturation of reconstruction error rate and the dependency of reconstruction error rate to the data size, of the newly proposed simulated annealing algorithm have been analyzed in depth. We have found out that the saturated reconstruction error rate is getting smaller as we increase the kinoform quantization level and the saturated reconstruction error rate of the same kino-form quantization level stays at the same level regardless of data size. And we also have found out that the threshold level should be increased as we increase data size to reconstruct data correctly from hologram ID pattern. Throughout this thesis...