摘要
为使特征提取更适合复杂矿石图像识别,提出并实现一种结合RGB颜色特征及其纹理特征映射的图像内容识别新方法,并将聚类方法应用于图像识别系统中。首先将图像分块,基于不同的颜色空间提取子块的纹理特征,并应用主成份分析进行纹理特征映射。然后提取图像的RGB颜色特征,每个子块的特征向量由上述2种特征组成。最后基于每个子块的特征向量应用Kmeans聚类方法对图像内容进行识别。实验结果表明,该方法能有效结合图像的纹理信息及其颜色构成和分布信息,具有较好的复杂矿石图像理解与识别的效果。
To make feature extraction more suitable for the recognition of complex ore image,a novel image content recognition method integrating the image RGB color feature and the corresponding texture was proposed.Firstly,the image is cut into small blocks,and the texture feature of each block is extracted,then the principle component analysis(PCA) is used to process texture mapping,afterwards,the RGB color feature is computed,.Then the feature vector of each sub-block is composed of the two kinds of features mentioned above.Finally based on the extracted feature vector,the image is recognized automatically by using Kmeans cluster method.Experimental results have shown that the proposed method has a good understanding and recognition of complex image content.by integrating the texture information into the color composition and color spatial information.
出处
《北京石油化工学院学报》
2010年第4期36-42,共7页
Journal of Beijing Institute of Petrochemical Technology
关键词
主成分分析
基于内容的图像分类
共生矩阵
图像纹理
RGB颜色
principle component analysis(PCA)
content-based image classification
co-occurrence matrix
image texture
RGB color