基于RGB空间的非经典K最近邻算法应用研究

为迅速、准确、无过多人工干预的进行图像分割，提出了一种K最近邻算聚类方法并将其应用于图像处理。与经典K最近邻算法在样本库中寻找最近邻点不同，该算法在待分割图像的RGB空间中寻找每一个像素点的K个最近邻点，参考所有像素点同最近邻点之间的平均距离，引入聚类阈值并对像素点的归属进行判断。对火焰图像的分割实验结果表明，在分割精度相接近的情况下，该算法的分割速度要快于其它几种常见算法。

APPLICATION RESEARCH OF NON-CLASSICAL K-NEAREST NEIGHBOR ALGORITHM BASED ON RGB SPACE

In order to segment the images quickly, accurately and without too much manual intervention, one K nearest neighbor clustering method was proposed and applied in image processing. Being different from the classical K-nearest neighbor algorithm, by which the nearest neighbor points were found in the sample set, the K nearest neighbors of each pixel in the RGB space of the image to be processed were found by the proposed algorithm. Referring to the average distance between all pixels and their K nearest neighbors, a clustering threshold associated with this distance was introduced and the attribution of each pixel was judged. The experimental results of flame image segmentation showed that when the segmentation accuracy was close, the segmentation speed of this algorithm was faster than that of other common algorithms.