A Stereo Image Segmentation Method Based on Interaction

Abstract

The invention provides an interactive stereoscopic image segmentation method. The original stereoscopic image segmentation method requires the user to paint consistent foreground and background lines in the left and right views, which is very tedious work. This method extends the method of single image segmentation to stereoscopic image segmentation, and only needs to paint the foreground and background lines in the left view to obtain consistent stereoscopic image segmentation results. First, the user interactively paints the foreground and background lines in the left view, and then samples the painted lines, image boundary points, and extracts the feature matching points of the left and right views, and uses these vertices and sampled line segments to construct the left view using the restricted Delaunay triangulation method Triangular mesh, and then construct an energy optimization equation to solve the grid with the right view consistent with the left view, so as to obtain the consistent coloring lines of the left and right views; then construct the feature vector to search the nearest neighbor pixels of each pixel, Finally, a non-local linear optimization model is constructed to perform image segmentation on the left and right views of the stereo image respectively.

Description

technical field [0001] The invention relates to an image segmentation method, in particular to an interactive stereoscopic image segmentation method. Background technique [0002] Image segmentation is the technology and process of dividing an image into several specific regions with unique properties and proposing objects of interest. Undoubtedly, image segmentation is one of the most important components in digital image processing. Image segmentation technology is mainly used in medical images, machine vision, target detection and other fields. Because the image segmentation problem is essentially an under-constrained problem. Therefore, most image segmentation methods require users to provide additional information in order to realize the solution to image segmentation. According to the way of providing information, the current mainstream image segmentation methods are mainly divided into the following two ways: (1) providing trimap, that is, the user is pre-divided i...

AI Technical Summary

Problems solved by technology

In order to maintain the consistency of stereoscopic image segmentation, users need to paint the foreground and background colors in the left and right views respectively. In order to reduce user interaction, Tasli et al. proposed a stereoscopic image segmentation method. Users only need to paint the foreground in the left view. and background lines, and then migrate the painted lines through ORB (Oriented fast and rotated BRIEF) feature detection, but the quality of this method depends entirely on the effect of feature detection. When the position feature points of the painted lines cannot be detected, the stereo image Segmentation results in an error

Images