Human face light re-adding method based on total variation model

Abstract

The invention discloses a human face light re-adding method based on a total variation model, comprising the following steps of: (1) establishing a multi-input logarithm total variation model; (2) inputting a human face image series of the same person into the multi-input logarithm total variation model; (3) decomposing the human face image series into a reflection component and a light illumination component by the multi-input logarithm total variation model; (4) inputting a human face light illumination image in a human face database into the multi-input logarithm total variation model; (5) decomposing the human face light illumination image in the human face database into a reflection component and a light illumination component by the multi-input logarithm total variation model; (6) combining the light illumination component in the step (3) and the light illumination component in the step (5) into a new light illumination component; and (7) combining the reflection component in the step (3) and the light illumination component generated by the step (6). According to the human face light re-adding method based disclosed by the invention, a human face image under extreme change light illumination can be formed; and single-image or multi-image input is supported, and the light illumination component and the reflection component can be accurately estimated.

Description

technical field [0001] The present invention relates to the technical field of face re-adding, in particular to a face re-adding method based on a total variation model. Background technique [0002] The so-called face re-lighting technology is to process the input face image and synthesize the face image under the lighting conditions specified by the user. Face re-lighting technology has a wide range of applications in video surveillance, media processing, digital art, public security investigation and other fields. For example, in the face recognition system, the face image has the same lighting conditions by re-lighting the face image, thereby improving the accuracy of face recognition. In image synthesis, face re-lighting technology can be used to coordinate the lighting of the face and the surrounding scene, so as to achieve a more realistic synthesis effect. [0003] The existing face re-lighting technologies mainly fall into the following categories: [0004] 1) Me...

AI Technical Summary

Problems solved by technology

[0012] The existing face re-lighting technology still has many shortcomings: for example, many methods can only support a single input image, and the information contained in a single image is limited. When estimating illumination, quotient image, reflectance, etc. In the shaded areas of the image often get wrong estimation results

The method based on the 3D face model requires a large number of input images of the same person in different poses, otherwise it is difficult to reconstruct an accurate 3D model

The method based on the polynomial model is difficult to deal with the situation where the illumination changes are very large

Images