Research on Human Image Understanding for the Coming Aged Society -Computer Eye Technologies to Watch People Kindly for Their Assistance-

通过计算机对人的图像的理解，对“亲切的眼睛”技术进行了基础研究，特别是对人的面部、眼睛、嘴巴、手、手指、身体姿势的理解方法进行了研究。本文以自然环境下人体活动的图像为基础，研究了人眼检测、人脸表情与识别、人体行走动画的物理建模与绘制技术。以及基于通用人体脊柱数据并使用新开发的连通球模型和精确脊柱模型的人体脊柱姿势识别。我们研究了各种医学图像处理方法，以生成精确的人体模型，从而精确分析和合成人体图像。本文还对手写数学公式的在线识别和大数据量三维数据的快速绘制进行了研究。

We conducted fundamental research on a "kind eye" technologies to watch people kindly for their assistance and protection by human image understanding by computers.Especially we examined methods for understanding human face, eyes, mouth, hand, fingers, and body postures. Based on human images acquired under natural human activity environment, we researched technologies for human eye detection, human face expression and recognition, physically based modeling and rendering of human walking animation. And human spine posture recognition based on a generic human spine data and using a newly developed connected sphere model and a precise spine model.We investigated methods for various medical image processing for the generation of very precise human models needed for accurate analysis and synthesis of human images. We also did research on the online recognition of handwritten mathematical formulas, and high speed rendering of large 3 dimensional data.