Collection And Reconstruction Of High Dimensional Optical Field Information

Traditional imaging technology can only record one dimension information of electromagnetic wave.When improving the system performance,key dimension information would be lost,such as the spectral camera which uses time to exchange for spectral resolution,the optical field camera which uses spatial resolution to improve angular resolution,etc.In this paper,a large depth of field camera based on birefringence is proposed,which can effectively expand the depth of field without sacrificing spatial resolution and color information.The innovations of this paper are as follows:1?A new type of photonic nano material(meta surface)is used to collect the optical field information.Meta surface material is a potential optical encoder with high plasticity,which can control the nano scale material structure and accurately control the multi-dimensional optical field information such as amplitude,phase and polarization.2?The system depth of field is increased by utilizing polarization and spectral dimensions.In this paper,the characteristics of high-dimensional optical field information are analyzed in detail.With the coupling ability given by the new optical encoder and the sparsity of high-dimensional information in nature,the information coupling extraction is almost completed in a lossless way.Compared with the traditional imaging methods,on the one hand,the polarization dimension information is effectively reused,which reduces the sacrifice of spatial resolution while expanding the depth of field;on the other hand,the sparseness of the spectral dimension is used to break through the limit of sampling frequency,and enough color information is retained while reusing the frequency domain.3?The decoupling reconstruction method based on learning is used.Compared with the traditional reconstruction algorithm,the algorithm based on learning is moreclose to the feature distribution of natural image,avoiding the limitation of artificial priori.Specifically,this paper uses the multi-scale full convolution neural network as the decoupling model,uses the actual calibrated system response function(PSF)to generate the training data set,and uses the system model to enhance the data set,so as to accurately simulate the imaging process and improve the reconstruction effect.The decoupling model is relatively simple and the reconstruction effect is very clear.In this paper,a large depth of field camera based on meta surface material is designed and fabricated.The system effectively expands the depth of field without sacrificing spatial resolution and color information.It is a successful practice in the cross field of meta surface and computational imaging.