Abstract: Auroral oval is very important for the study of magnetic structures and energy storage. This paper proposes a segmentation algorithm of aurora image based on region growing. Firstly, it raises the concept of original image pretreatment according to the characteristics of aurora image. Based on the gray characteristics of pretreatment image, it also needs to select a group of the representative seed points in object region. Secondly, adopting the OTSU and determining the best adaptive segment threshold conveniently based on the features of neighborhoods’ pixels gray, this method replaces the system errors which caused by the traditional region growing method, then determines region-growing stopping conditions by the method of regional difference of gray value. Finally, by using the method of mathematical morphology, it eliminates the voids and discontinuous phenomenon which caused by the noise in the regional growth. Some experimental data and results of image segmentation are presented and proven this algorithm improves the quality of image segmentation.

Key words: auroral oval, region-growing, seed points, optimal segmenting threshold, image segmentation

摘要： 极光卵对于研究磁层结构和能量存储是非常重要的。提出一种基于区域生长的极光图像分割算法。首先根据极光图像的特点，对原始图像进行预处理，根据预处理后图像的灰度特性，选取一组能正确代表目标区域的种子像素；其次，在分析像素邻域灰度特性的基础上，采用最大类间方差法求得自适应最佳阈值，从而代替传统区域生长算法手动选取阈值时造成的系统误差，再采用基于区域灰度差的方法，制定出区域生长的停止条件；最后，利用数学形态学的方法进行分割后修正，消除了由于噪声而造成生长后的区域中出现的空洞和不连续现象。实验结果表明，相对于已有的极光卵分割方法，基于区域生长的方法改善了图像的分割质量。

关键词: 极光卵, 区域生长, 种子点, 最大类间方差法, 图像分割