基于BoF模型的多特征融合纹理图像分类

Texture image classification based on BoF model with multi-feature fusion

针对特征词袋(BoF)模型缺乏空间和几何信息,对纹理图像内容表达不明显等问题,提出一种基于BoF模型的多特征融合纹理分类算法。将灰度梯度共生矩阵(GGCM)和尺度不变特征转换(SIFT)融合特征作为纹理图像的区域特征描述,通过动态权重鉴别能量分析进行最优参数特征选择,并用BoF量化纹理特征,使用支持向量机对图像进行训练和预测,得出分类结果。实验结果表明,本文算法对有旋转扭曲的纹理、边缘模糊纹理、有光照变化的纹理及杂乱纹理等均能取得较好的分类效果,相对于传统BoF模型及凹凸划分(CCP)方法等算法在UIUC纹理库上的分类正确率均有不同程度的提高,平均分类正确率分别提高12. 8%和7. 9%,说明本文算法针对纹理图像分类具有较高的精度和较好的鲁棒性。

The obvious shortcomings of bag of feature（ BoF） model are lack of spatial and geometric information in image representation,and poor description of the content of texture image. To solve these problems,we proposed a texture image classification method based on BoF model with multi-feature fusion. The method fuses gray gradient co-occurrence matrix（ GGCM） and scale-invariant feature transform（ SIFT） as the basic feature description of texture image,uses a dynamic weight to identify energy analysis for the optimal parameter feature selection,quantifies texture feature by BoF,then applies support vector machine to train and predict the image,and finally obtains the classification results. The experimental results show that the proposed method has better performance of texture classification into rotated texture,twisted texture,edge fuzzy texture,light changing texture,messy texture,etc. The average classification accuracy of the proposed method on the UIUC texture database increases by 12. 8% and 7. 9% respectively compared with the conventional BoF model and concave-convex partition（ CCP） methods,which indicates that the proposed method has higher accuracy and better robustness for texture image classification.