Single image patch reconstruction method based on undirected graph learning model

Abstract

The invention discloses a single image patch reconstruction method based on an undirected graph learning model. The method comprises the steps of performing multi-view rendering on three-dimensional models in an existing three-dimensional model set to obtain a training image set; representing the patch by using an undirected graph learning model, establishing an undirected graph initialization network consisting of image coding and camera view angle estimation, obtaining camera view angle parameters corresponding to the image, and projecting the initial patch according to the camera view angleparameters obtained by estimation to obtain undirected graph initial characteristics; establishing an undirected graph LSTM network structure and an undirected graph CNN network structure, performingfeature updating on the undirected graph model, and mapping each node of the undirected graph to coordinates in a three-dimensional space to obtain each vertex position of the triangular patch; establishing an undirected graph initialization network and an undirected graph updating network loss, and performing multi-stage training on the neural network; and performing three-dimensional reconstruction on the input image by using the trained network to obtain a final patch model.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to a method for reconstructing a single image patch based on an undirected graph learning model. Background technique [0002] Three-dimensional reconstruction is to restore the three-dimensional shape of the object contained in the image from the image using a specific technology. However, this task is a morbid problem, because the self-occlusion problem will inevitably appear in the image, the information provided is limited, and other prior information needs to be added to complete it. [0003] In fact, some 3D reconstruction methods have been proposed in academia, among which the 3D reconstruction method based on visual cues is a method of 3D modeling of objects in the image directly based on the computer vision features in the image and guided by the physical knowledge of the real world. . Such as literature 1: Andrew, Alex M. "Shape from Shading, edited b...

AI Technical Summary

Problems solved by technology

Since the simple neural network structure is difficult to directly obtain the topological structure of the triangular patch, only the parameterized triangular patch model can be obtained in advance, and the quality of the model parameter representation will affect the quality of the reconstructed patch, and this method There are many category restrictions, and it is difficult to adapt to diverse model categories, so it is necessary to use a neural network structure designed for triangular patches to complete triangle patch reconstruction

Images