Human body scene image eigen-decomposition and relighting method and device

Abstract

The invention provides a human body scene image eigen-decomposition and relighting method and device, and the method comprises the steps: making a human body model through employing Clo3d, carrying out the light tracing through employing blender, and carrying out the rendering, and obtaining a training data set; pre-training a normal prediction network by using a single image; taking the original image and the normal graph obtained through prediction as input of an image eigen-decomposition network, and obtaining a reflectivity graph, a shadow graph and spherical harmonic illumination parameters through prediction; re-illuminating the original image by using the new spherical harmonic illumination parameter; and finally, decomposing a single picture into a reflectivity picture and a shadow picture, so the application of picture relighting, 3D model synthesis, model surface texture reconstruction and the like can be achieved. The shadow of the input picture can be removed in the field of three-dimensional reconstruction, and the surface texture quality of a three-dimensional reconstruction model is improved.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a method and device for eigendecomposition and re-illumination of a human body scene image. Background technique [0002] Image eigendecomposition is a classic visual image processing problem, which involves decomposing a single image into multiple images representing the intrinsic properties of the image, and is an underdetermined problem. The common image eigendecomposition method decomposes the image into albedo map, shading map, and further, normal map, light map and BRDFs can be obtained from a single image. Image eigendecomposition has many applications, such as 3D model synthesis, model surface texture reconstruction, and relighting. [0003] Image decomposition methods can be divided into two categories: optimization-based image eigendecomposition and deep learning-based image eigendecomposition. The image eigendecomposition based on optimization relies on stron...

AI Technical Summary

Problems solved by technology

However, the image intrinsic method based on optimization is only suitable for image decomposition in simple scenes, and the image decomposition effect in complex lighting scenes is not good.

Images