广义局部二值模式在彩色图像检索中的仿真

Color Image Retrieval Simulation via Generalized Local Binary Pattern

局部二值模式是一种经典的图像特征生成算法,为进一步加强局部二值模式特征在模式识别中的鲁棒性,提出一种广义局部二值模式特征。广义局部二值模式在传统局部二值模式基础上增加了幅值、方差以及梯度等信息,在彩色图像中将广义局部二值模式组成特征向量,通过多模态局部敏感哈希对彩色图像的三通道特征进行哈希处理得到最优相似矩阵。对于需要检索的图像则计算其与图像库中相应特征的哈希距离。通过在国际公开的Corel图像库和华盛顿大学图像数据研究组搜集供图像检索数据库进行仿真,仿真效果表明建议的广义局部二值模式在彩色图像检索中具有较强的竞争力。

Local binary pattern is a classical image feature generation algorithm.To further enhance the robustness of local binary pattern features in pattern recognition,a generalized local binary pattern feature is proposed in this paper.The generalized local binary mode adds information such as amplitude,variance and gradient on the basis of the traditional local binary mode.In color images,the generalized local binary pattern was composed into the feature vector,and the multi-modal local sensitive Hash was applied to the color image.The three-channel feature was hashed to obtain the optimal similarity matrix.For the image that needs to be retrieved,the hash distance of the corresponding feature in the image library was calculated.The simulation results show that the proposed generalized local binary mode has strong competitiveness in color image retrieval by collecting images for the image retrieval database in the internationally published Corel image library and the University of Washington image data research group.