基于二元共生模式的图像检索系统

Binary co-occurrence Pattern Based Image Retrievalsystem

基于内容图像检索方法大多仅对某个类型的图像具有较好的性能,无法适用于不同类型的图像,提出一种基于二元共生模式与灰度共生矩阵的基于内容医学图像检索方法。首先,从原图像提取中心对称局部二元模式以获得图像的局部信息;然后,使用灰度共生矩阵提取局部模式图中不同方向与距离的共生像素对。该方法使用共生像素对的模式频率,其鲁棒性优于基于直方图的局部模式图。最终基于2个纹理数据集与一个医学图像数据集进行检索实验,结果表明该方法具有较好的检索性能,并具有较好的时间效率。

The existing content based image retrieval algorithms only show good performance for a special category of images, and are not suitable for different kinds of images, a binary co-occurrence pattern and gray level co-occurrence matrix based content based medical image retrieval algorithm is proposed. Firstly, the center symmetric local binary co-occurrence pattern is abstracted from the original image to obtain the local information of the image; then, the gray level co-occurrence matrix is used to abstract the co-occurrence pixel pairs with different distances and different directions from the local pattern image. The pattern frequency of the co-occurrence pixel pairs is used in the proposed method, so that the proposed method has better robustness than the histogram based local pattern image. Finally, two texture image datasets and a medical image dataset based experimental results show that the proposed method has better retrieval performance and better time efficiency.