摘要
根据协同神经网络 ,提出了一种新的控制参数———序矢量 ,并着重研究了其在图像检索方面的相关性质 .在此基础上 ,针对商标图像检索 ,提出了一种高效的基于元素的仿射不变性度量方法 .实验表明 ,该检索算法抗噪、抗缺损能力强 ,同时对于平移、旋转、缩放具有不变性 .
A new concept of control parameter, order vector, based on SNN was proposed. More specifically, its associate properties for image retrieval was studied. Based on the properties, an efficient element based affine invariance similarity measurement was proposed for retrieval of trademark images. The experiments indicate that the retrieval method is good in against noisy and occluded images, and the method has affine invariance in translated, rotated and scaled images.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
2002年第5期327-331,共5页
Journal of Infrared and Millimeter Waves
基金
国家自然科学基金 (批准号 69772 0 0 2 )资助项目~~
关键词
图像检索
相似性度量
仿射不变性
协同神经网络
Image retrieval
Similarity measurement
Affine invariance
synergetic neural net（SNN）.
