Prior algorithm-based facial expression image occlusion restoration method

Abstract

The invention discloses a prior algorithm-based facial expression image occlusion restoration method, comprising the following steps: S1, processing an original facial image; S2, performing model training; S3, restoring the shielded facial expression image; S4, collecting facial motion units; S5, judging the expression of the restored image; and S6, performing simulation transformation on the secondary restored image. According to the prior algorithm-based facial expression image occlusion restoration method, after the neural network is restored, the facial expression image can be judged, andwhether the restoration effect is reasonable and accurate or not is judged; the accuracy of the repairing effect is greatly improved, and the prior algorithm-based facial expression image occlusion restoration method conforms to the law of facial muscle movement of the human face and is more real; and when a facial expression image is subjected to simulation transformation, transformation coordinate parameters can be modified by adjusting amplitude parameters, and coordinates of key feature points are also transformed along with the transformation coordinate parameters, namely, the mouth corner opening degree, so that the expression transformation amplitude can be obtained. The prior algorithm-based facial expression image occlusion restoration method improves the accuracy of operation.

Description

technical field [0001] The invention relates to the technical field of facial expression restoration, in particular to a priori algorithm-based occlusion restoration method for facial expression images. Background technique [0002] In recent years, with the rapid development of information technology and the wide application of computers, facial expression occlusion restoration and its recognition technology have been widely used in human-computer interaction, public safety monitoring, daily monitoring, psychological analysis, entertainment industry, etc. Face recognition technology faces a large number of rich application requirements, but also faces different challenges, such as lighting, occlusion, posture, environment and angle and other factors. Occlusion is very important among many factors affecting the performance of face recognition, and it is also an unavoidable factor. Therefore, an efficient and accurate method is needed to repair facial occlusion expression ima...

AI Technical Summary

Problems solved by technology

Usually these repair methods cannot achieve accurate repair. With the introduction of more and more neural networks, there are more and more methods for facial expression repair.

In the past, GAN network, PCA+SVM method, sparse representation method (Sparse representation based classification, SRC) and CNN method, all of them will lead to the accuracy of expression restoration is not ideal and not realistic enough in the partial occlusion problem. The expressions are not very real and coherent

This will also lead to poor results in the field of face recognition and a decrease in recognition accuracy.

[0005] For the implementation of the above method, only the facial expressions that cover the face are repaired, and no judgment algorithm after repair is added. Because only the discriminator in the network model is used to correct the accuracy of the image, there is no correction for the repaired image. Judging and identifying, which makes the restored expression not conform to the law of facial muscle movement to some extent, making the restoration effect not good

At the same time, there will be some errors with the original expression in the process of expression restoration, and eventually the expression of some expressions will not conform to the expression of the original expression image

Images