应用支持向量机的纹理分类

提出了一种使用离散余弦变换(DCT)进行特征提取的应用支持向量机的纹理分类算法,并将文章中的算法与 KIM K I 等提出的不进行先期特征提取而直接将纹理图像送入支持向量机进行训练分类的算法进行比较。结果显示,文章中的算法可以取得更为准确的分类结果,能够大大降低分类错误率,并且分类结果受参数变化的影响很小。由此说明,在使用支持向量机进行纹理分类的过程中,准确的先期特征提取十分必要。

Texture classification by support vector machines

The arithmetic of texture classification by support vector machines (SVM) was investigated, using discrete cosine transform (DCT) to extracting external features. It was unlike the arithmetic proposed by KIM K I in which the gray-level values of raw pixels were fed to SVM directly without any external feature extraction. It is shown that the external feature extracting is necessary in the application of support vector machines in texture classification because the arithmetic proposed in this paper can classify the texture more exactly, reduce the error rate greatly and almost get rid of the influence of parameter changing.