Adaptive Parameters Estimation for Light Field Reconstruction using Shearlet Transform

The research of light field can provide the VR (Virtual Reality)/AR (Augmented Reality) products' more realistic immersion effect and bring revolutionary development to the VR/AR industry. The light field reconstruction is one of the most important technologies in light field. The methods to reconstruct the light field can generally be divided into the time domain and the frequency domain. The advantage of reconstructing the light field from frequency domain is that it can eliminate the effect of occlusion on the light field reconstruction to a certain extent. Although many reconstruction algorithms in frequency domain has been put forward, their performance on different data sources strongly depends on pre-set fixed parameters. There is a great need for a method that can adaptively adjust parameters in dealing with various sources of data to improve this situation. This paper utilizes the correlation between the parameters to improve the efficiency of light field reconstruction using Sheartlet Transform and proposes an adaptive parameter estimation method to adjust the parameters to be suitable for all types of light fields. Simulation results illustrates that the PSNR is obviously higher than that of the original method, which verify the adaptability of the proposed parameter estimation method of the shearlet light field reconstruction to different light fields.