中文摘要：

米粒组织是太阳表层对流运动所产生的一种形态特征.由于米粒强度分布不均匀以及边缘比较模糊,使得采用传统的基于强度和梯度阈值的方法来准确地识别它们变得困难.因此,本文提出了一个基于相位一致性的米粒识别算法.选用中国科学院云南天文台抚仙湖的新一代太阳真空望远镜（New Vaccum Sloar Telescope,简称NVST）的高分辨活动区观测资料来展示算法的识别过程,并根据识别结果分析了米粒的形态特征.选取了2个目前已存在的米粒识别算法来验证所提算法的准确性和有效性,实验结果表明所提算法能够有效地提取那些低对比度的米粒特征.同时为了检验算法对阈值的响应程度,分别选取了3组阈值来检验不同阈值情况下的差异性.结果证明所提算法对阈值响应不敏感.为了进一步检验所提算法获取统计结果的准确性,对米粒的直径、强度、形状以及分形维数进行统计.统计结果很好地符合了已有文献的结论,这也进一步验证所提算法的准确性和健壮性,能够用于进一步的科学研究.

英文摘要：

Solar granulation is a plasma convective phenomenon from the solar convection zone to the photosphere, which appears as a cellular pattern on the solar surface. Morphological characteristics and evolution process of granules can help us to better understand the mechanism of the convective motion and the relationship to solar magnetic field. Therefore, an accurate identification method is crucial to study the mechanism and relationship. However, the traditional identification methods based on intensity- and gradient-threshold are very difficult to identify granules because the intensity distribution of granules is non-uniform and their boundaries appear to be blurred. In this paper, we proposed a new method based on phase congruency to identify granules and analyzed their morphological characteristics. The identification procedure includes three steps： （1） obtaining phase congruency feature from an original image; （2） obtaining binary image of phase congruency feature; （3） morphological filter to extract granule shapes. We selected two high-resolution image sequences taken from the New Vacuum Solar Telescope （NVST） at Fuxian Solar Observatory （FSO） to illustrate the identification procedure. For evaluating the accuracy of the proposed method, two traditional methods based on intensity threshold and marked watershed were compare to our method. Furthermore, we also selected three different thresholds to inspect the influence. The experiment result demonstrates that our proposed method is effective and accurate, and insensitive to those selected threshold. This demonstrates that our proposed method can be used to study granules evolution and their physical mechanism. Using the method, a total of 165694 and 108279 granules were identified from the first and the second data sets, respectively. The diameter of each granule was first obtained, and then their distributions were exhibited. We found that the distributions of granules from the two data sets hold two obvious peaks, implying that the gra