Illumination detection method and apparatus for face image, and device and storage medium

A device, comprising:

• one or more processors;
• and
• a memory, configured to store one or more programs;
• wherein the one or more processors, when loading and running the one or more programs, are caused to:
• acquire a face image to be detected and a three-dimensional face mesh template;
• acquire a reconstructed face mesh model by deforming the three-dimensional face mesh template based on the face image;
• determine, based on deformation positions, in the reconstructed face mesh model, of key feature points in the three-dimensional face mesh template, brightness of feature points, corresponding to the key feature points, in the face image;
• and
• determine illumination information of the face image based on a corresponding relationship between predetermined brightness of the key feature points and illumination thereof, and the brightness of the feature points, corresponding to the key feature points, in the face image, wherein the corresponding relationship between predetermined brightness of the key feature points and illumination is acquired by training with history face images of which illumination information has been determined, and brightness of feature points, corresponding to the key feature points in the three-dimensional face mesh template, in the history face images,
• wherein prior to acquire the face image to be detected and the three-dimensional face mesh template, the one or more processors are further caused to:
• acquire history face images and the three-dimensional face mesh template, and determining history illumination information of the history face images;
• acquire reconstructed history face mesh models by deforming the three-dimensional face mesh template based on the history face images;
• determine, based on deformation positions, in the reconstructed history face mesh models, of the key feature points in the three-dimensional face mesh template, brightness of feature points, corresponding to the key feature points, in the history face images;
• and
• determine the corresponding relationship between the predetermined brightness of the key feature points and the illumination thereof based on the history illumination information of the history face images and the brightness of the feature points, corresponding to the key feature points, in the history face images;
• wherein the one or more processors are further caused to:
• for each history face image, determine a corresponding relationship between initial brightness of the key feature points and the illumination thereof based on the history illumination information of the history face image and the brightness of the feature points, corresponding to the key feature points, in the history face image;
• and
• acquire the corresponding relationship between the predetermined brightness of the key feature points and the illumination thereof by processing the corresponding relationships between the initial brightness of the key feature points and the illumination thereof, corresponding to each of the history face images, using a least square method.

A non-transitory computer-readable storage medium storing a computer program therein, wherein the program, when loaded and run by a processor, causes the processor to:

• acquire a face image to be detected and a three-dimensional face mesh template;
• acquire a reconstructed face mesh model by deforming the three-dimensional face mesh template based on the face image;
• determine, based on deformation positions, in the reconstructed face mesh model, of key feature points in the three-dimensional face mesh template, brightness of feature points, corresponding to the key feature points, in the face image;
• and
• determine illumination information of the face image based on a corresponding relationship between predetermined brightness of the key feature points and illumination thereof, and the brightness of the feature points, corresponding to the key feature points, in the face image, wherein the corresponding relationship between predetermined brightness of the key feature points and illumination is acquired by training with history face images of which illumination information has been determined, and brightness of feature points, corresponding to the key feature points in the three-dimensional face mesh template, in the history face images,
• wherein prior to acquire the face image to be detected and the three-dimensional face mesh template, the program further causes the processor to:
• acquire history face images and the three-dimensional face mesh template, and determining history illumination information of the history face images;
• acquire reconstructed history face mesh models by deforming the three-dimensional face mesh template based on the history face images;
• determine, based on deformation positions, in the reconstructed history face mesh models, of the key feature points in the three-dimensional face mesh template, brightness of feature points, corresponding to the key feature points, in the history face images;
• and
• determine the corresponding relationship between the predetermined brightness of the key feature points and the illumination thereof based on the history illumination information of the history face images and the brightness of the feature points, corresponding to the key feature points, in the history face images;
• and
• wherein the program causes the processor to:
• for each history face image, determine a corresponding relationship between initial brightness of the key feature points and the illumination thereof based on the history illumination information of the history face image and the brightness of the feature points, corresponding to the key feature points, in the history face image;
• and
• acquire the corresponding relationship between the predetermined brightness of the key feature points and the illumination thereof by processing the corresponding relationships between the initial brightness of the key feature points and the illumination thereof, corresponding to each of the history face images, using a least square method.