二级分区下颜色融合纹理的刑侦图像检索

多数刑侦图像检索算法缺乏图像局部特征的提取,从而导致检索精确度较低。对此,提出一种新的二级分区下颜色融合纹理的刑侦图像检索算法。该算法采用二级分区将图像划分成多个局部区域;提取局部区域的颜色自相关图特征和双树复小波结合多方向多参数的灰度共生矩阵的纹理特征;以串行的方式融合局部区域特征。使用核主成分分析法(Kernel Principle Component Analysis,KPCA)进行特征选择,以欧式距离作为相似性度量实现检索。通过在两个刑侦图库的实验结果表明,该算法的平均精确度均高于传统的刑侦图像检索算法,具有良好的检索效果。

CRIMINAL INVESTIGATION IMAGE RETRIEVAL BASED ON COLOR AND TEXTURE FUSED UNDER TWO-STAGE PARTITION

Most criminal investigation image retrieval algorithms lack the extraction of image local features,which leads to low retrieval accuracy.To solve this problem,we proposed a novel criminal investigation image retrieval algorithm based on color and texture fused under two-stage partition.We used two-stage partition to divide the image into several local regions,extracted the color auto-correlogram features of the local region and the texture features of the dual-tree complex wavelet combined with the gray level co-occurrence matrix of multi-direction and multi-parameter,and fused the local region features in a serial way.Then we adopted the kernel principal component analysis(KPCA) to select features,and used Euclidean distance as similarity measure to achieve retrieval.The experimental results of two criminal investigation image databases show that the average accuracy of the algorithm is higher than that of the traditional algorithm,and it has a good retrieval effect.