Light field semantic segmentation method and system, electronic terminal and storage medium

Abstract

The invention provides a light field semantic segmentation method and system, an electronic terminal and a storage medium. The method comprises the steps that: a reference view angle is selected froma camera plane so as to perform light field sampling; the superpixel set of the reference view angle is obtained based on a superpixel segmentation algorithm, and reprojection is performed on the superpixel set of the reference view angle to obtain superpixel sets of other view angles corresponding to the reference view angle; focal length fusion is carried out on a plurality of images with different re-focusing depths in a focusing stack, and the superpixel sets to which pixels belong are voted; semantic analysis is performed on the images in the focusing stack based on a neural network algorithm to obtain the semantic classification of each superpixel set; and the semantic classifications of all the superpixel sets in the focusing stack are summarized, and voting is performed, so that aunique semantic category number corresponding to each superpixel set is determined. The method is no longer limited by depth information loss caused by projection transformation, and effectively recognizes occlusion so as to perform correct category prediction on the pixel points of an occluded object.

Description

technical field [0001] The present invention relates to the field of light field technology, in particular to a method, system, electronic terminal and storage medium for semantic segmentation of light field. Background technique [0002] A light field is a commonly used representation of light in computer photography and refers to the amount of light passing through each point in each direction. Semantic segmentation is one of the basic tasks in computer vision. This task takes a two-dimensional image as input, uses a visual algorithm to segment the object area in the image, and recognizes the content, that is, while ensuring the continuity of the image domain, it is Each pixel of the image determines the semantic category it belongs to. [0003] Light field semantic segmentation refers to assigning semantic classification to each ray in the scene under the setting of the light field. Two-dimensional images of different depths are semantically segmented, and the two-dimen...

AI Technical Summary

Problems solved by technology

[0005] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a light field semantic segmentation method, system, electronic terminal and storage medium, which are used to solve the problem that the traditional two-dimensional image domain segmentation algorithm is often limited by the depth caused by projection transformation. Due to the lack of information, there are complex occlusions in most real scenes, and the algorithm cannot effectively identify occlusions, so it is impossible to make correct category predictions for the pixels of occluded objects.

Images