Resolution-Enhancement for an Integral Imaging Microscopy Using Deep Learning

Abstract

A novel resolution-enhancement method for an integral imaging microscopy that applies interpolation and deep learning is proposed, and the complete system with both hardware and software components is implemented. The resolution of the captured elemental image array is increased by generating intermediate-view elemental images between each neighboring elemental image, and an orthographic-view visualization of the specimen is reconstructed. Then, a deep learning algorithm is used to generate maximum possible resolution for each reconstructed directional-view image with improved quality. Since a pretrained model is applied, the proposed system processes the images directly without data training. The experimental results indicate that the proposed system produces resolution-enhanced directional-view images, and quantitative evaluation methods for reconstructed images such as the peak signal-To-noise ratio and the power spectral density confirm that the proposed system provides improvements in image quality.