基于改进的ISODATA算法彩色图像分割

如何对彩色图像中的目标进行快速、精确的有效分割是计算机视觉和图像分析的重点和难点。提出了一种基于区域的彩色图像分割方法。该方法首先选择合适的彩色空间，提取出图像中的每个像素点的颜色、纹理、位置等综合特征，形成特征向量空间；在特征空间中，运用改进的ISODATA算法自适应地确定初始聚类数目和聚类中心，然后对图像进行聚类和区域分割，最后抽取出图像区域的特征，并与相类似的方法进行了比较实验。实验结果表明，该方法能够产生较好的分割效果及较快的分割速度，适合于基于图像区域检索系统，具有较好的应用价值。

Color Image Segmentation Based on Improved ISODATA

How to fast, accurately and effectively segment objects in the color images is the key point in the computer vision and image analysis. This paper introduces a method of region-based color image segmentation. This method first extracts color, texture, and location features for each pixel form integrative feature vectors by selecting suitable color space to form the feature space. In the feature space, the initial cluster center and the number are determined by the improved ISODATA algorithm adaptively, then an image is clustering and separated into regions. Finally, the features of regions are extracted. The experimental results and the comparision with the similar approach are provided. Experimental results show the proposed method has high segmentation speed and good sementation results, and it is fit for region-based image retrieval system and has better application values.