Re: Re: Counting circles (digital image processing)
- To: mathgroup at smc.vnet.net
- Subject: [mg64548] Re: Re: Counting circles (digital image processing)
- From: "Jens-Peer Kuska" <kuska at informatik.uni-leipzig.de>
- Date: Wed, 22 Feb 2006 05:58:30 -0500 (EST)
- Organization: Uni Leipzig
- References: <dte258$rde$1@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
Hi, and huuu K-means and all of its variants work on the distribution of the pixel values and do abolut nothing with the geometrie of the objects shown in the picture ... The OP need the number of extended geometric objects, in the worst case that are white circles on an black background and the distribution has a peak at 0 and one at 1 and nothing more ... K-means will find the threshold to binarize the image if it is nor already, but it will never find connected pixels. The image processing package has a connected component labeling *and* a distance transform (don't know which one) and it should be simple to count the circles and get the radius. The only problem may be, that the images are extrem noisy or the circle shape can be seen only when one looks at the gradients of the image and not onto the intensity, but in both cases K-means help nothing. Regards Jens "Maarten van der Burgt" <maarten.vanderburgt at icos.be> schrieb im Newsbeitrag news:dte258$rde$1 at smc.vnet.net... | | Hallo, | | I think the function KMeans[ ] in the Digital Image Processing package is | the clustering function you are looking for. | Have a look in the User`s Guide section 7.5 Segmentation by Clustering. | | Maarten | | | ||--------+---------------------------------> || | "Jens-Peer Kuska" | || | <kuska at informatik.uni-l| || | eipzig.de> | || | | || | 18/02/2006 08:49 | || | | ||--------+---------------------------------> | >------------------------------------------------------------------------------------------------------------| | | | | | To: mathgroup at smc.vnet.net | | | cc: | | | Subject: [mg64548] Re: Counting circles (digital image processing) | | >------------------------------------------------------------------------------------------------------------| | | | | | Hi, | | label the connected components in the binarized image. | For the radius I would suggest to use a distance transform on | the binary image and the maximum value of the distance transform | in every single unconnected component is your radius | | Regards | Jens | "Alexander" <beginning.physst at mail.ru> schrieb im | Newsbeitrag news:dt488e$2hk$1 at smc.vnet.net... || Dear Mathgroup! || || Suppose we have a picture with a set of white | circles on a black || background. || How can we count circles and measure their | radius using mathematica || system??? || || Thanks for your answers! || Alexander. || | | | | | |