1

So I have a visible-light image and a set of fluorescent microscopic picture which consist of 2 channels, Green and Red. In the sequence visible-light, green channel and red channel

Green channel signal happens on the whole cell's body. The red channel signal mostly happen on the membrane of the cell. There are 4 combinations possible: cell with no signal, cell with red and green signal, cell with red signal and cell with green signal.

The white channel's segmentation leads to a failure because of the low image quality and resolution. Green channel's segmentation is much easier because the a simple BW threshold can give us a good BW mask for the cells. However the red channel image is more tricky as I will explain below. A BW image generated from Green and Red channel is attached below to show the difference.

green channel Red Channel

The objective of the experiment is: 1. Count the number of the cells with red signal. 2 For cells has red signals, check if it has green signal. I don't care about the cells without red signal.

But as you can see after BW thresholding, the red signal sometimes return solid cells; but more likely return a hollowed circular object and often incomplete. Is there a way to find out the region that covered by, or likely covered by the circular object? So I can find out the average green channel intensity within the region after registered the images.

I have attached the original red channel file. Thank you very much for your help!

Tianyu Yang
  • 87
  • 4
  • how sophisticated do you need to be? – bla Jul 29 '15 at 23:49
  • @bla Is that possible you elaborate the question? What do you mean by sophisticated. – Tianyu Yang Jul 30 '15 at 00:30
  • 1
    [This sophisticated](http://stackoverflow.com/questions/18972932/image-segmentation-based-on-edge-pixel-map/19510366#19510366)? – Shai Jul 30 '15 at 04:30
  • @Shai Whoa.. This is definitely an eye opening for me. I can barely understand them. Just consider me a undergrads who suddenly dived into image processing without knowing anything before hands. Is there simpler methods? Also seems like you are an expert on this, is that possible you can tell me some topics I can research on in terms of edge detection? I have some other datasets like this white light image in this problem that require cell segmentation. I tried marker based watershed but facing difficulty when the image quality is bad. – Tianyu Yang Jul 30 '15 at 23:04
  • @Shai I solved this particular problem with brutal force. But Another question just arrived and all my attempts has failed. It is a similar cell segmentation but this time I am dealing with visible light and bad image quality. In the link you sent me mentioned a "Python Edge Classifier" that generate a Edge Pixel Map. I think it is quite useful but I can't find a way to do it except very high level description on Google, and I don't have enough reputation to comment. I am wondering if you know how to do it. You can just give me a list of topics, I can research them out myself. Please help! – Tianyu Yang Aug 20 '15 at 00:04

0 Answers0