项目名称: 视觉媒体的结构感知处理与分析模型研究
项目编号: No.61502541
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 计算机科学学科
项目作者: 苏卓
作者单位: 中山大学
项目金额: 22万元
中文摘要: 结构感知滤波模型是针对现有的边缘感知滤波方式的新拓展,并从根本上突破了传统滤波方法的局限性。最近,结构感知滤波成为视觉媒体处理与分析方法研究中的热点,原因在于它能有效获取视觉媒体数据中的边界、轮廓、形状和细节信息。然而,当前的边缘或结构感知滤波均存在一定的局限性,限制了其在视觉媒体中的应用。本项目致力于研究结构感知滤波的理论与实践方法,主要研究:(1)显式稀疏边缘控制的优化滤波模型。在稀疏梯度范数最小化的优化框架中,结合显式的稀疏边缘信息,构造具有结构感知特性的滤波结果。(2)基于局部全变分的自适应核回归模型。在核回归理论中,结合局部全变分理论对结构和纹理信息的表达能力,构建具有结构感知效果的滤波方法。(3)基于结构感知滤波的视觉媒体处理与与分析质量优化策略。通过探索结构感知滤波与视觉媒体内容处理过程的结合方式,发掘提高处理质量的新方法。最终的研究成果将在数字家庭媒体平台上加以推广应用。
中文关键词: 图像滤波;图像视频的结构分析;图像增强;图像视频编辑;图像缩放
英文摘要: Structure-aware filtering model is an extension to current edge-preserving smoothing models, and essentially breaks the limitations in the traditional filters. Recent yeas, the structure-aware filter becomes a significant way to the processing and analysis of the visual media, since it effectively separetes the media data to obtain the boundaries, contours, shapes and details. However, there exists some defects in the state-of-the-art models. In this project, we devote ourselves to explore the fundamental theories and the practical methods of the new structure-aware filter. The main points are as follows. (1) Explicit sparse edge control optimization filtering model. It exploits the sparse edges to construct the structure-aware property in the sparse gradient-norms minimization framework. (2) Local total variation based adaptive kernel regression model. Under the kernel regression theory, it exploits the powerful structure-texture expression ability of the local total variation to establish the structure-aware filter. (3) The schemes for the visual media processing and analysis by structure-aware filtering. It studies the effective combination patterns for raising the visual media processing and analysis quality. All the research achievements will be promoted to the visual applications in the digital home environment.
英文关键词: Image Filtering;Structure Analysis for Image/Video;Image Enhancement;Image/Video Editing;Image Retargeting