分水岭matlab仿真,正确使用Matlab分水岭算法对细胞进行分割

我've run into a ton of trouble trying to get the watershed algorithm properly working on my images. In the various tutorials online they always seem to use images that are just as complicated/fuzzy and so I'我不确定我的错误 . 我've already made a couple of spotty posts on this, but wanted to really clarify and ask generally. That said, I'm使用的图像如下：

但是，当我尝试应用其中一个分水岭算法时：

imshow(RGB,[]);

gray_img = rgb2gray(RGB);

tophat_filter = imtophat(gray_img, strel('disk', 10)); %Read into this

level = graythresh(tophat_filter);

BW = im2bw(tophat_filter,level);

imshow(BW)

BW = bwdist(BW) <= 3;

imshow(BW)

bgn_remove = bwareaopen(BW,8); %remove background noise

D = -bwdist(~bgn_remove); %Read into this

D(~BW) = -Inf;

L = watershed(D);

figure;

imshow(L,[]);

figure;

imshow(label2rgb(L))

clean_img = im2bw(L,0.001);

figure;

imshow(clean_img,[]);

它似乎永远不会起作用 . 无论出于何种原因，它确定每个单元格包含一堆较小的单元格：

我试图通过使用 BW = bwdist(BW) <= 3; 聚集segmeneted组件来解决这个问题，以便图像不是碎片化的：

如第一张图片中所示，应该有3个细胞，虽然分水岭确实识别出左侧的两个不同的细胞 - 它的注册量超过应有的程度(即使在聚集后) . 在我尝试过的所有内容之后，我没有取得多大进展，所以任何帮助或建议都会非常感激 .

在完成整个分水岭程序后，我最终得到了最大值：