Automatic elimination method of artifact generated in face image synthesis

Abstract

The invention discloses a method for automatically eliminating virtual shadows in human face image synthesis. In the process of synthesizing a new face image from two or more face images, through a set of specific synthesis formulas, a large number of ghost images and ghost images caused by traditional synthesis methods are eliminated. When the traditional synthesis method has errors in the calibrated face feature points, these errors are amplified, causing virtual images and double images in the synthesis results, reducing the image quality, and seriously making local areas of the face unrecognizable. This method improves on this problem. No matter whether there is an error in the input feature data, the synthesized face image can not only integrate the facial features of the original image, but also be clear and smooth without ghosting, and the visual effect is good.

Description

technical field [0001] The invention relates to the fields of computer image processing and computer vision, in particular to an automatic elimination method for ghost images in human face image synthesis. Background technique [0002] With the rapid development of computer software and hardware technology, especially computer vision theory and image processing technology, various theoretical and practical application methods in the specific field of face image detection, recognition and fusion are also becoming more and more abundant. Face recognition and synthesis have a very wide range of application prospects, from movie special effects to medical search; from public safety to auxiliary teaching; from virtual conferences to home entertainment, etc. It can be seen everywhere. [0003] The general flow of face image processing is: first preprocess the original image, then detect the face area, identify the features of the face, and then synthesize the face according to spe...

AI Technical Summary

Problems solved by technology

However, if there is an error in the facial feature point data obtained in the previous stage of face synthesis—face recognition, the error will continue to be amplified during the fusion process, and a large number of virtual edges will appear on the edge of the final synthesized face. shadow, which reduces the image quality and visual effect

Objectively speaking, the feature points obtained by the current face recognition method inevitably have different degrees of error, which makes the probability of facial ghosting and ghosting in the fusion stage very high.

Images