基于OTSU分割与K-均值聚类的MPEG-7主颜色提取算法

AN EXTRACTION ALGORITHM FOR MPEG-7 DOMINANT COLOUR BASED ON OTSU SEGMENTATION AND K-MEANS CLUSTERING

针对图像检索系统,提出一种MPEG-7主颜色提取的改进算法。首先通过OTSU阈值对图像进行分割,得到图像的前景和背景,然后应用K-均值聚类算法分别提取前景和背景的主颜色,最后综合前景和背景的颜色特征进行图像检索。改进算法既减少了颜色特征维数,又综合了颜色的空间分布信息,并且对图像的旋转和绝对位置不敏感。对比实验结果表明此方法具有较好的检索效果。

An improved algorithm for extracting dominant colour in MPEG-7 aiming at image retrieval system is proposed. Firstly the foreground and background of an image is obtained through image segmentation using OTSU threshold,then K-means clustering algorithm is applied to get the main colour of both foreground and background,finally integrated colour features of foreground and background is used for image retrieval. The improved method reduces the dimensions of feature vector,incorporates spatial distribution information,and is insensitive to rotations and absolute locations. Contrasting experimental results show that this method has a better retrieval effect.