摘要
研究粘连现象严重的胃上皮细胞图像准确分割问题,针对胃结构排列状态复杂,分割起来非常困难。为了能够准确地区分出图像的细胞核、细胞浆和背景区域,用改进的传统Mean Shift算法,选择高斯函数为核函数,有效确保了Mean Shift算法的收敛性,避免了图像分割的过分割,并设置核函数的带宽为可变性,方便对粘连现象不同的图像分割。同时将改进的分割算法与松弛迭代分割、分水岭分割和传统的Mean Shift分割算法进行了仿真效果对比。实验结果表明利用改进的算法分割的效果较好,改进了传统的Mean Shift分割算法等的过分割问题,提高了分割速度和准确性,在错综复杂的胃上皮细胞图像分割中具有可行性和有效性。
Image segmentation is very difficult due to the severe adhesion of gastric epithelial cell image.In order to well distinguish nucleus,cytoplasm and background of gastric images,this paper improves the traditional Mean Shift algorithm and chooses Gaussian function as kernel function.The algorithm effectively ensures the convergence of Mean Shift algorithm and avoids over-segmentation problem of image segmentation.The band width of the kernel function is set variability,which facilitates image segmentation of different extend adhesion.Then the improved segmentation algorithm is compared with relaxation segmentation,watershed segmentation and the traditional Mean Shift segmentation algorithm through simulation results.It is proved by the experimental results that the segmentation algorithm has better effect,improves the over-segmentation problem of the traditional Mean Shift segmentation algorithm and increases the segmentation rate,which is feasible and effective in the complex image segmentation of gastric epithelial cells.
出处
《计算机仿真》
CSCD
北大核心
2011年第8期242-245,共4页
Computer Simulation
基金
江西省科技支撑计划项目:CAD图形数字水印系统研究(赣财教[2008]212号)
关键词
核函数
松弛迭代
分水岭
图像分割
Nuclear function
Relaxation
Watershed
Image segmentation