基于局部灰度聚类的高斯分布拟合模型

发布时间：2018-08-03 17:58

【摘要】：针对高斯分布拟合模型对初始轮廓敏感的问题,提出一个基于局部灰度聚类的高斯分布拟合模型.新模型根据图像局部像素灰度聚类特点,采用灰度偏移场和一个分片常量函数共同拟合图像的局部灰度均值,实现了图像全局信息和局部信息的有机结合,使轮廓可以从任意初始位置向目标边缘演化,最后收敛在边缘上.新模型采用一种快速有效的数值方法实现,水平集函数在整个演化过程中不必重新初始化,活动轮廓演化速度得到显著提高.实验结果表明,本文算法能够在不同的轮廓初始化情况下获得准确的分割结果.
[Abstract]:In order to solve the problem that Gao Si distribution fitting model is sensitive to initial contour, a Gao Si distribution fitting model based on local gray clustering is proposed. According to the characteristics of image local pixel gray clustering, the new model uses the gray level offset field and a piecewise constant function to fit the local gray mean value of the image, and realizes the organic combination of the global and local information of the image. The contour can evolve from any initial position to the edge of the target, and finally converge on the edge. The new model is implemented by a fast and efficient numerical method. The level set function does not have to be reinitialized during the whole evolution process, and the evolution speed of the active contour is improved significantly. Experimental results show that the proposed algorithm can obtain accurate segmentation results under different contour initialization conditions.
【作者单位】：东北大学信息科学与工程学院;
【基金】：国家自然科学基金资助项目(61273078) 中央高校基本科研业务费专项资金资助项目(N140403005)
【分类号】：TP391.41

【相似文献】