视觉显著性纹理--色彩特征融合的图像目标分类

Vision saliency texture-color feature fusion basedobject classification

针对图像目标分类,提出了一种显著性纹理特征。考虑到显著目标图像在纹理特征表征上的优势,在目标显著性图像提取的基础上进一步提取视觉显著性纹理特征。进而将该视觉显著性纹理特征同HSV色彩特征进行融合,形成图像目标融合特征,输入至后端分类器中进行分类。多类别的交叉实验证明,基于该融合特征的目标分类方法能够较为准确的对图像目标进行分类,在SIMPLIcity图像数据集上平均分类正确率达到84.84%,在Corel图像集上平均分类正确率为85.05%,优于基于单一分类特征的图像分类方法。

In order to improve the performance of the image classification, a novel saliency texture feature is proposed. Considering the advantage of the saliency map for representing the texture information, the saliency texture feature is extracted based on the saliency map. This feature is further fused with the HSV color feature, generating a fused image feature which is inputted into the classifier. Results of cross-over experiments demonstrate that the fused feature works better than the compared counterpart and has the ability to correctly recognize the image objects. The imageelassification precision rates of the proposed method in SIMPLicity and Corel5k databases achieved 84.84% and 85. 05%, respectively.