Generating method of face image and generating device of face image

Abstract

The invention discloses a generating method and device of a face image. The method comprises: obtaining an RGB three-channel color image of a real face, and generating a face position rectangle frameand facial feature point coordinates according to the same; obtaining PCA (Principal Component Analysis) appearance parameters and PCA expression parameters of a 3DMM (3-Dimensional Morphable Model) face model and an affine transformation matrix, of projecting the three-dimentional face to the RGB three-channel color image according to the face position rectangle frame and facial feature point coordinates to reconstruct a three-dimensional model of the face; generating color information, which corresponds to each sampling point in a three-dimentional lattice which is of the face and is obtained by three-dimentional reconstruction, according to the RGB three-channel color image and the reconstructed three-dimentional model of the face; generating a four-channel RGBA face image and face positioning point coordinates on the basis of the color information, the PCA appearance parameters, the PCA expression parameters, light ray material and face angle information; and generating the face image according to a pre-collected general scene picture labeled with a face position and the four-channel RGBA face image. Existing color images can be converted into near-infrared (NIR) images, and collection costs of the near-infrared images are reduced.

Description

technical field [0001] The invention relates to the technical field of face detection and face recognition, in particular to a method for generating a face image and a device for generating a face image. Background technique [0002] Face detection and face recognition are topics of great concern in the field of computer vision. With the advancement of deep learning algorithms, the performance and effect of face detection and recognition algorithms based on deep neural networks are far superior to traditional methods. However, due to the high dimensionality of face image information, training deep neural networks requires a large number of image samples. In many cases, labeled image samples are required. Collecting samples from the real world and marking them often requires a lot of manpower and material resources. In the category of near-infrared images, very limited public samples are available. Contents of the invention [0003] The present invention aims to solve at l...

AI Technical Summary

Problems solved by technology

However, due to the high dimensionality of face image information, training deep neural networks requires a large number of image samples. In many cases, labeled image samples are required. Collecting samples from the real world and marking them often requires a lot of manpower and material resources. In the category of near-infrared imagery, very limited public samples are available

Images