项目名称: 融合目标感知与对比度的图像和视频显著性检测技术研究
项目编号: No.61502220
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: 林晓
作者单位: 洛阳师范学院
项目金额: 20万元
中文摘要: 本项目针对显著性检测时易产生边界模糊、内容提取不完整、准确性不足和时空不一致等问题,研究融合目标感知与对比度的图像和视频显著性检测技术。首先,在对比度和目标边界的基础上,研究融合区域颜色对比度和显著边界的图像显著性检测算法,克服图像显著性检测结果中边界模糊和内容提取不完整的缺点;其次,在前景目标获取的基础上,研究目标感知的图像显著性检测模型,使得在复杂背景下仍能够得到准确、边界清晰和内容完整的图像显著度图;最后,在视频运动等动态特征的基础上,研究基于动态对比度的视频显著性检测算法,使得视频显著性检测结果准确且符合时空一致性的要求。本项目对于图像和视频显著性检测技术的研究为后续图像和视频处理如缩放、压缩、目标识别、分割等打下坚实的理论基础,促进相关技术的快速发展,具有较强的研究和应用价值。
中文关键词: 目标感知;对比度;显著性检测
英文摘要: In order to resolve fuzzy boundaries, incomplete contents and inaccuracy in image saliency detection, and inconsistence of spatio-temporal in videos, we research the saliency detection combining target perception with contrast of images and videos. Firstly, on the basis of known contrast and target boundary, we research the image saliency detection combining regional color contrast with saliency boundary, to overcome the shortcoming of fuzzy boundaries and incomplete contents. Furthermore, we propose a novelty saliency detection based on target perception in complicated background, which is able to obtain an accurance saliency map with clear boundary and complete content. Especially for inconsistence of spatio-temporal videos, we gain a saliency detection algorithm based on dynamic contrast of videos, leading to higher accurancy and continutity of spatio-temporal results. This research play an important role in resizing, compression, target recognition and segmentation. It is advantageous to promote the development of these technologies, and which is useful in many related applications.
英文关键词: Target Perception;Contrast;Saliency Detection