Re: Segmenting touching particles

ImageAnalyst <imageanalyst@xxxxxxxxxxxxxx> wrote in message <3a374c23-315f-429e-9e32-9625afff101e@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>...
On Sep 22, 9:34?am, "Jasper Engel" <jaspereng...@xxxxxxxxxxx> wrote:
Hello,

Im trying to count different sized particles in a microscopic image (http://img186.imageshack.us/i/hfets4d4.jpg/) following the ricegrain example on this website. Most particles are counted properly. However, particles close in each others vicinity are counted as one big particle (http://img199.imageshack.us/i/bigparticle.jpg/). ?Any advice on how to improve upon this would be greatly appreciated!

Kind regards,

Jasper

--------------------------------------------------------------------------------------------------
Jasper:
Why don't you try marker controlled watershed segmentation to separate
them? MATLAB has a demo for this in the image processing toolbox.
Alternatively just find the mean or mode area, and go through your
particles and if the area is more than 2 times the mean count it as 2
particles (instead of 1), if it's more than 3 count it as 3, more than
4 count it as 4 etc.
Regards,
ImageAnalyst

Thank you very much! A watershed on the distance matrix works pretty well. However, I have another question. When comparing the watershed picture and the original picture some particles are a larger and some are smaller. Are there any techniques to get error estimates for the measured particle area and particle count?

Regards,

Jasper
.

Relevant Pages

  • Re: Segmenting touching particles
    ... Why don't you try marker controlled watershed segmentation to separate ... MATLAB has a demo for this in the image processing toolbox. ... Alternatively just find the mean or mode area, ...
    (comp.soft-sys.matlab)
  • ICALAB toolbox with Matlab
    ... I need help for ICALAB for image processing toolbox of Matlab. ... to separate some mixed images, ...
    (sci.image.processing)
Loading