基于改进小波包变换的纹理图像检索方法

The Method of Texture Image Retrieval on Improved Wavelet Packet Transform

通过对基于小波包变换的纹理图像检索方法的改进,提出了一种新的纹理图像检索方法.首先用小波包变换对图像进行分解,然后利用小波包分解得到子图,从中选取m个最大的子图能量值构造纹理特征向量来描述纹理图像的纹理信息,并在此基础上利用这些纹理特征向量进行纹理图像检索.实验结果表明,该方法能够对纹理图像进行很好的分类和检索.

A new method for texture image retrieval is presented which improves the method of texture image retrieval on wavelet packet transform. After decomposing texture image by wavelet packet transform, we construct texture feature eigenvectors by the m most dominant energy values of subgraph to describe the texture information of texture image, and then the result of image retrieval can be obtained by using these texture feature eigenvectors. The experiment shows that classification and retrieval over texture image have been achieved successfully.