基于细分的几何逼近与可视媒体处理中的非线性理论与方法

项目名称： 基于细分的几何逼近与可视媒体处理中的非线性理论与方法

项目编号： No.61472466

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 檀结庆

作者单位： 合肥工业大学

项目金额： 62万元

中文摘要： 本项目研究目标是以非线性理论与方法为基础平台, 对几何逼近和可视媒体内容处理中的若干关键技术展开研究，以丰富和发展计算几何、计算机图形学、模式识别和视频处理、分析与理解中的基本理论与方法。具体的研究内容包括：设计一种能够有效表示复杂视频场景目标的非线性模型；探索视频图像超分辨率重建中的混合向量连分式插值方法；去除视频中高密度脉冲噪声的非线性方法；设计并实现快速、有效的视频修复算法；构造动态均匀的三点三重插值细分格式；构造插值与逼近相统一的三点三重细分格式；构造保形细分格式；探究在非均匀控制顶点下的动态插值细分格式；构造和研究具有分形性质的细分格式；研究将曲线细分方法推广到任意拓扑结构的曲面细分格式；研究四重细分格式的相关性质和造型能力；探究曲面造型细分格式和细分曲面的光顺性等。

中文关键词： 视频超分辨率重建；视频修复；视频目标跟踪；细分逼近；非线性方法

英文摘要： The objective of this project is to conduct the research on some key problems related to the geometric approximation and visual media processing by means of nonlinear theory and methods and to develop and enrich the basic theory and methods for computational geometry, computer graphics, pattern recognition as well as video processing, analysis and understanding. The contents include, but are not limited to the followings: the design of the nonlinear models which effecively represent the complex visual object; the exploration of the blending vector valued interpolation methods by continued fractions applied to the video super resolution reconstruction; the nonlinear methods for the removal of high density pulse noises in videos; the design and implementation of fast and effective video inpainting algorithms; the construction of dynamic uniform three-point ternary interpolation subdivision schemes; the construction of the three-point ternary subdivisions by combining interpolation with approximation; the construction of the shape preserving subdivisions; the exploration of the dynamic interpolation subdivisions with nonuniform control points; the study of the subdivisions with fractal properties; the study on the extention of curve subdivisions to surface subdivisions with arbitrary topological structure; the study on the properties of the quaternary subdivisions; the study on the fairing of subdivision curves and surfaces.

英文关键词： Video super-resolution reconstruction;video inpainting;video object tracking;subdivision approximation;nonlinear methods