Single-image large-pose three-dimensional color face reconstruction method based on UV position map and CGAN

Abstract

The invention discloses a single-image large-attitude three-dimensional color face reconstruction method based on a UV position map and a CGAN, and the method comprises the steps: generating a three-dimensional point cloud model through employing UV position recording, complementing an incomplete face through employing a network based on the CGAN design, and finally obtaining a complete color three-dimensional face model. designing coding-decoding network, generating and recording a two-dimensional UV position map of complete three-dimensional face information from an original RGB image, and then remolding a three-dimensional face from the two-dimensional UV position map by using a convolutional neural network; and then, considering the self-shielding condition of the large posture of the face, and complementing the deficiency of the UV texture image by designing a special conditional generative adversarial network. According to the method provided by the invention, higher reconstruction precision and more texture details can be realized, a more complete and real three-dimensional face model can be obtained especially in large-pose face image reconstruction application, and the method has stronger robustness in coping with complex environmental factors.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular, a single-image large-pose three-dimensional color face reconstruction method based on UV position map and CGAN Background technique [0002] Biological feature is a kind of information feature that has been widely concerned and used recently, and the reconstruction technology of the corresponding model is also constantly developing with the change of social needs. The rich feature information contained in the face makes it an important carrier for person identification, expression recognition, age and gender judgment, so the processing of face information has always been an important research topic in the field of computer vision. However, the face information that can be retained in a two-dimensional image is very limited, and it will be affected by the shooting angle, object occlusion, and lighting angle. The recently popular 3D reconstruction technology has also been greatly impr...

AI Technical Summary

Problems solved by technology

[0003] The existing 3D face reconstruction technology can use a 3D model obtained from a single image, but due to the large angle of the face in the image, the reconstruction result has a large error, and the lack of complete surface texture of the model reduces the realism

The present invention provides a single-image large-pose three-dimensional color face reconstruction method based on UV position map and CGAN. In a single image, a large number of faces are invisible due to self-occlusion of large-pose faces, resulting in a decrease in the accuracy of three-dimensional face reconstruction and The end result is missing a lot of face color textures

Images