Design of an Optical Transfer Function Classifier based on Machine Learning and Deep Learning for Optical Scanning Holography

Abstract

Optical Transfer Function in Optical Scanning Holographic (OSH) System describes the mathematical model of hologram generation frequency domain. Here a deep learning feature vector extractor is used for combining the features of the hologram to the classifiers. The classification learning is done with the regression-based machine learning models. This system works as the pupil function predictor for the generated hologram. The training is done with the given dataset for different types of pupil functions. The extracted features of the hologram determine the model prediction for pupils used and then classification of OTF is performed. The accuracy measure for different learning algorithms has been analyzed and the Ensemble Adaboost classification algorithm shows best accuracy results for the prediction of the pupils used in OSH. This classification algorithm gives an average prediction accuracy of 97.75%