利用分块截短编码法实现图像信息鉴别

Block Truncation Coding Method to Achieve Image Information Identification

图像聚类是图像信息检索鉴别的关键技术,在图像信息系统中,图像的检索效率是至关重要的,在基于内容的图像鉴别中,如果能利用低级别的可视特征进行高效图像聚类,将大大提高图像信息的检索鉴别精度。文章分别利用色矩与BTC算法(分块截短编码)来提取图像的特征并进行图像特征K均值聚类,从而验证两种图像信息鉴别方法的效率,实验结果表明,分块截短编码(BTC)算法可以实现较高效率的图像鉴别。

Image clustering is the key technology for identification and searching of image information. The retrieval efficiency of image is critically important in pictorial information system. If we use low-level visual features for efficient image clustering in the content-based image identification, we will greatly improve the accuracy of image retrieval and identification. The Color Moment method and the Block Truncation Coding （BTC） method are used to extract color features respectively for clustering features using K-Means. The efficiencies of these two methods for the identification of image information are validated by experiment. The experimental result shows that the BTC （Block Truncated Coding） method to extract the image features for image clustering in order to identify image with higher efficiency.