真实感人脸网格模型的纹理映射研究
本文选题:真实感 切入点:人脸模型调整 出处:《武汉理工大学》2012年硕士论文
【摘要】:随着计算机图形学技术的不断发展和完善,虚拟现实技术在影视动画制作、广告设计、游戏娱乐、生物特征识别等方面的应用日益广泛。在这众多的研究领域中,创建逼真的特定人脸三维模型一直是一个极富挑战性的课题。由于人脸面部特征的复杂性,如何保证模型与人脸的特征匹配,并且对模型实现具有真实感的纹理映射是关键问题,也是本文的研究重点。具体研究内容包括: 1.光照模型建立。从真实感图像绘制流程出发,基于OpenGL光照模型实现了给场景中的模型添加光照效果,提高了三维模型显示的真实感。通过调节场景中环境光、散射光、镜面光和发射光各自的参数比例,将它们分别计算后叠加起来,形成最终的光照效果。 2.二维人脸特征提取。采用基于肤色的人脸识别算法,先通过相似度计算和阈值分割对图像进行二值化处理,根据处理后的结果判断出人脸的大致范围,再通过人工交互界面修正算法局限性导致的人脸范围判断失真的地方,最后提取出人脸的轮廓及特征信息。 3.三维模型调整。提出了基于比例变换的模型调整算法,首先将三维模型从上至下分割成n块,求出每一块的横坐标最大和最小的点,从而计算出该块的宽度,根据宽度以及整个模型的高度算出该块的比例系数。类似地,求出提取的人脸轮廓每一行的比例系数。依据这些比例系数来插值调整三维模型上点的三维坐标值,使其与二维人脸轮廓保持一致。 4.纹理图像到三维模型的映射。为了更具真实感,要保证将纹理图像上的关键特征与三维网格模型上相应的特征位置准确对应。本文采取了简单约束纹理映射的方法来控制三维特征点与二位纹理之间的映射关系,即两点定位后计算关键点的位移差值,根据这个差值改变原来的纹理空间坐标与三维空间坐标的对应关系,以此来实现模型的特征点精确匹配的效果。 基于VC++6.0和OpenGL图形库设计并实现了三维人脸模型的纹理映射系统。该系统能够对输入的单幅人脸照片进行人脸检测与特征提取,并通过手工修正及自适应的模型调整来得到目标人脸的三维模型,并对其进行了约束纹理映射和光照调节,使得结果真实感效果更强。
[Abstract]:With the development and perfection of computer graphics technology, virtual reality technology has been widely used in film and television animation production, advertising design, game entertainment, biometric recognition and so on. It has always been a challenging task to create realistic 3D face models. Because of the complexity of facial features, how to ensure the matching between the models and the features of human faces. It is a key problem to realize realistic texture mapping of the model, and it is also the focus of this paper. The specific research contents include:. 1. The illumination model is established. Based on the OpenGL illumination model, the illumination effect is added to the model of the scene, and the reality of the 3D model is improved. By adjusting the ambient light and scattering light in the scene, the illumination model is constructed. The ratio of the parameters of the specular light and the emitting light is calculated separately and superimposed to form the final illumination effect. 2. Two-dimensional face feature extraction. Face recognition algorithm based on skin color is adopted. Firstly, the binarization of image is processed by similarity calculation and threshold segmentation, and the approximate range of face is judged according to the processed results. Then the artificial interactive interface is used to correct the distortion of the face range caused by the limitation of the algorithm. Finally, the contour and feature information of the face are extracted. 3. 3D model adjustment. A model adjustment algorithm based on proportional transformation is proposed. Firstly, the 3D model is divided into n blocks from top to bottom, the maximum and minimum points of each block are obtained, and the width of the block is calculated. The proportional coefficients of the block are calculated according to the width and the height of the whole model. Similarly, the ratio coefficients of each line of the extracted face contour are calculated. The three-dimensional coordinate values of the points on the 3D model are interpolated according to these proportional coefficients. Make it consistent with two-dimensional face contours. 4. Mapping texture images to 3D models. In order to ensure that the key features on the texture image are accurately matched with the corresponding feature positions on the 3D mesh model, a simple constrained texture mapping method is adopted to control the mapping relationship between 3D feature points and binary textures. That is to say, the displacement difference of the key points is calculated after two points are located, according to which the corresponding relation between the original texture spatial coordinates and the three-dimensional spatial coordinates is changed to achieve the accurate matching effect of the feature points of the model. Based on VC 6.0 and OpenGL graphics library, a texture mapping system for 3D face model is designed and implemented. The 3D model of the target face is obtained by manual modification and adaptive model adjustment, and the constrained texture mapping and illumination adjustment are carried out to make the result more realistic.
【学位授予单位】:武汉理工大学
【学位级别】:硕士
【学位授予年份】:2012
【分类号】:TP391.41
【参考文献】
相关期刊论文 前10条
1 尹宝才,胡永利,程世铨,谷春亮;基于形变模型的人脸建模及其应用综述[J];北京工业大学学报;2003年03期
2 程日彬;周明全;李春龙;;基于二维图像的三维人脸建模技术[J];计算机工程与应用;2006年03期
3 赵向阳,杜利民;人脸合成中模型的平滑调整和逼真的纹理映射[J];计算机辅助设计与图形学学报;2004年11期
4 苏志勋;赵元棣;曹俊杰;;约束纹理映射的自适应方法[J];计算机辅助设计与图形学学报;2009年12期
5 卢章平;丁立军;戴立玲;;基于分类的纹理映射方法综述[J];江苏大学学报(自然科学版);2006年S1期
6 赵晓刚;惠延波;;特定三维人脸建模方法综述[J];计算机与数字工程;2009年05期
7 涂意;赵群飞;卢芳芳;;基于单张人脸图片和一般模型的三维重建方法[J];计算机应用研究;2009年10期
8 王成章;尹宝才;孙艳丰;胡永利;;改进的基于形变模型的三维人脸建模方法[J];自动化学报;2007年03期
9 龚勋;王国胤;;基于特征点的三维人脸形变模型[J];软件学报;2009年03期
10 署光;姚莉秀;陈鲁;杨杰;;基于一般模型的单幅人脸照片三维重建[J];上海交通大学学报;2009年03期
相关硕士学位论文 前4条
1 郭洋;基于神经网络的单张照片三维人脸建模[D];北京邮电大学;2011年
2 李增忠;纹理映射技术的研究[D];西安电子科技大学;2005年
3 张洁;基于纹理映射真实感图形的研究[D];西安电子科技大学;2007年
4 陈婵娟;真实感三维人脸贴图技术研究[D];西北大学;2010年
,本文编号:1657059
本文链接:https://www.wllwen.com/wenyilunwen/guanggaoshejilunwen/1657059.html