A Light Field Reconstruction Method

Abstract

The invention discloses a light field reconstruction method, which can keep structural information among all dimensions of high-dimensional data, more effectively carries out sparse representation on the data, lowers a storage space of a dictionary and greatly improves the quality of an image of a reconstructed light field. A light field atom is represented as a tensor structure in a training dictionary, a plurality of small dictionaries are simultaneously trained according to each mode of a tensor, and the small dictionaries are combined into a big dictionary through a Kronecker product.

Description

technical field [0001] The invention belongs to the technical field of computational photography, and in particular relates to a light field reconstruction method. Background technique [0002] Compared with traditional color or grayscale images, light field image sets can provide a more accurate and reliable description of the scene, and can improve the performance of various tasks in the field of computer vision. In recent years, more and more light field cameras have emerged, one is to directly obtain the light field through the design of optical devices or adding the rearrangement of sensing pixels; the other is to obtain part of the light field data, and the data will be obtained Perform post-computational processing to obtain the reconstructed light field. [0003] In SIGGRAPH 2013, Marwah et al. proposed a new type of compressed light field camera, which is realized by inserting a mask in the optical path and reconstructing it with compressed sensing theory. This me...

AI Technical Summary

Problems solved by technology

However, the light field is a kind of high-dimensional data, and the vectorization operation destroys the structural information of the high-dimensional data. Therefore, the data cannot be sparsely represented, the storage space of the dictionary is relatively large, and the quality of the reconstructed light field image is poor.

Images