摘要
针对时域和频域纹理特征的优点和互补性,提出一种结合局部二值模式(LBP)和Brushlet域系数统计特征的自适应纹理图像检索方法。利用Brushlet变换得到各个子带的能量作为频域特征,提取图像的LBP直方图作为空域特征,并采用改进的Canberra距离进行度量,使用闭环反馈实现图像的自适应检索。实验结果表明,与LBP方法和Brushlet方法相比,该方法的平均检索率分别提高8.93%和18.66%。
By considering advantages and complementary of spatial and frequency textures features,an adaptive texture image retrieval method is presented combined with Local Binary Pattern（LBP） feature and Brushlet domain coefficient statistical feature.Frequency feature is described as energy feature of every Brushlet subband.Spatial feature is LBP histogram.Similarity between images is measured by improved Canberra distance.Further,closed-loop feedback is introduced to adjust weights adaptively for image retrieval.Experimental results show that average recall rate of this method based on fused features is 8.93% and 18.66% higher than LBP method and Brushlet domain method respectively.
出处
《计算机工程》
CAS
CSCD
2013年第2期233-236,共4页
Computer Engineering
基金
国家自然科学基金资助项目(60802061)
智能感知与图像理解教育部重点实验室开放基金资助项目"图像检索中相关反馈技术研究"(IPIU012011004)
关键词
图像检索
BRUSHLET变换
局部二值模式
闭环反馈
空域特征
变换域特征
image retrieval
Brushlet transform
Local Binary Pattern（LBP）
closed-loop feedback
spatial domain feature
transform domain feature
