基于小波变换的图像纹理特征提取方法及其应用

A Method of Image Texture Feature Analysis Based on Wavelet Decomposition and its Application

针对瓷砖表面色差的在线检测问题,提出了一种基于小波变换的图像纹理特征提取方法。瓷砖图像经过预处理后,对图像各通道进行二层小波分解并提取各细节子图的能量特征。该能量信号融合了颜色和纹理的信息,将其作为区分不同色号瓷砖的特征量,并由最小距离分类器进行分类决策。实验结果表明,对比已有的研究中多采用的颜色直方图分布法,该方法能更好地反映颜色的空间分布信息,满足用户需求。

For detecting color difference of ceramic tiles on-line,a method of analyzing the textural features of image is put forward by using the wavelet decomposition. First the image of ceramic tile is pretreated. Then two level wavelet decomposition is used on the ceramic tile image, and the energy feature of each subgraph is extracted. The energy feature will be used as a feature matrix, and it can be carried out the elassfication according to minimum distance classifier. Experiments show that this mehtod can reflect spatial relationship of color more effectively, when contrast with the color histogram which is used as a traditional method.