摘要
传统人脸三维图像重构优化方法受到光线等外界环境的影响,导致人脸三维图像重构精度较低。为此,基于激光扫描技术提出了一种人脸三维图像重构的优化方法。利用激光扫描人脸图像,将人脸特征点汇总组成形状向量,并得到纹理表达式,利用小波变换将参数矢量划分为若干个子带,并对各子带内的噪声进行去除,实现人脸特征点定位。构建三维稀疏形变重构模型,手动标定三维人脸特征点,并采用改进后的投影法去除人脸三维图像中的噪声,获得更加真实且符合要求的人脸三维图像。将人脸柱状展开判定肤色点,并进行非肤色区域修正,在肤色区域进行人脸三维图像纹理恢复优化,从而完成人脸三维图像重构优化。实验结果表明:主观观察下,设计方法得到的优化图像可以实现侧视多角度和多模式图像的有效还原,三组图像面部特征全部恢复,重构效果较好,统计数据结果表明,设计方法重构后特征点定位误差不超过0.024 8,图像的质量有较大提升。
The traditional optimization methods for facial 3D image reconstruction are affected by external factors such as lighting,resulting in lower accuracy of facial 3D image reconstruction.Therefore,an optimization method for facial 3D image reconstruction based on laser scanning technology is proposed.Using laser scanning of facial images,facial feature points are summarized into shape vectors and texture expressions are obtained.The parameter vectors are divided into several sub bands using wavelet transform,and the noise within each sub band is removed to achieve facial feature point localization.Build a three-dimensional sparse deformation reconstruction model,manually calibrate the three-dimensional facial feature points,and use an improved projection method to remove noise from the facial threedimensional image,obtaining a more realistic and satisfactory facial three-dimensional image.Expand the facial column to determine skin color points,and perform non skin color area correction.Optimize the texture restoration of the facial 3D image in the skin color area,thus completing the optimization of facial 3D image reconstruction.The experimental results show that under subjective observation,the optimized images obtained by the design method can achieve effective restoration of side view multi angle and multi-mode images.The facial features of all three sets of images are restored,and the reconstruction effect is good.The statistical data results show that the feature point positioning error after reconstruction by the design method does not exceed 0.0248,and the image quality is greatly improved.
作者
吴薇
WU Wei(Nanchang Institute of Technology,Nanchang 330044,China)
出处
《激光杂志》
北大核心
2025年第12期187-192,共6页
Laser Journal
基金
江西省自然科学基金项目(No.20214BAB001203)。
关键词
激光扫描
人脸图像预处理
三维稀疏形变模型
小波变换
肤色修正
laser scanning
facial image preprocessing
three dimensional sparse deformation model
wavelet transform
skin tone correction
