摘要
针对目前基于内容的图像检索技术中低级特征无法准确全面地描述高级语义的问题 ,本文提出了一种基于多级图像描述模型的渐进式图像内容理解 .该图像描述模型在不同层次上对图像内容进行分析和提取 ,实现了图像内容的全方位描述 ,从底层向高层的过渡是渐进式的图像理解过程 .特别是从视觉感知层到目标层 ,体现了图像低级特征与高级语义之间的过渡 .本文给出了一种基于先验知识的上下文驱动的目标理解算法 ,实现了图像语义的提取 .作为一个应用实例 。
An new method for progressive image content understanding based on multi-level image description model is proposed, aiming at overcoming the considerable gap between low-level image features and high-level image semantics in the field of image retrieval. In the proposed image content description model, image contents are analyzed and extracted in different levels, reaching at omnidirectional image content description. In addition, the transition from low level to high level is exactly a progressive image understanding. A new algorithm for object understanding is proposed, which is based on pre-knowledge and is context-driven, in order to extract image semantics. As a practical instance, discussion about combining the proposed method into content-based image retrieval is also given.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2001年第10期1376-1380,共5页
Acta Electronica Sinica
关键词
图像内容理解
图像描述模型
图像检索
Algorithms
Content based retrieval
Feature extraction
Mathematical models
Object recognition
Semantics
