基于纹理谱及其空间特征的图像检索

Image retrieval based on texture spectrum and its spatial feature

针对传统纹理谱描述符维数较高且忽略了空间特征的问题,一方面进一步考虑邻域中心像素与其它像素间的灰度关系,改进了中心对称局部二值模式纹理谱描述符,另一方面基于灰度共生矩阵提出了纹理谱基元共生矩阵的概念,并据此来提取纹理谱基元空间特征。通过不同图像库进行实验,结果表明,结合所提取的空间特征,可明显地提高传统纹理谱描述符的检索性能,而且所提出的描述符以较低的维数取得了较好的检索效果。

In order to solve the problems of high dimension and lack of spatial feature of the traditional texture spectrum descriptors, the paper defines the local patterns based on the relativity of central pixels and center-symmetric pixels which are different from local binary pattern （LBP） and center-symmetric local binary pattern （CS-LBP）, and then, introduces texture spectrum co-occurrence matrices （TSCMs） based on the gray-level co-occurrence. After that, the statistics calculated from TSCMs are used as the spatial feature. The methods mentioned above were tested on three different databases and the results prove that the retrieval performance of the traditional texture spectrum descriptors can be improved markedly in image retrieval if the spatial feature is considered. The results also show that the new approaches presented in the paper have better performance than the traditional descriptors.