I tried various functions in matlab but non of them could distinguish the boundary of the cross sections in the image.is any one to help?
so thanks
As i found, bmp image is cross section of objects. Before apply the processing algorithm, you should be aware of objective. Is this image is used to count the object or size of object finding or something else?
If it is used for counting, then first step is to extract the object from background (edge detection). Second, convert in binary image. Find out the correlation factor between image and template of one object (scale the template if required). Find the centroid of each object. Count the centroids.
If the objective is not this, then mention your purpose.
I suppose the question is how to segment individual cells or what ever the image contains? This can be difficult problem in general, but for your case for example Maximally Stable Extremal Regions (MSER) could be a good alternative. With reasonably new Matlab and Image processing toolbox you can use "detectMSERFeatures" to extract MSER regions.If the toolbox is missing etc try to search net for free VLfeat toolbox that contains also MSER. An example script for Matlab 2014a is attached that yield result(s) shown in the given result image.
The preprocessing the input may turn out to be quite critical step (remove noise, but keep edges etc) and you may have to tweak that to get the approach work for "all" inputs. While the segmentation works reasonably you may have to join overlapping results together etc...
If this approach is not good enough maybe you can try out some graph cut based segmentation methods...
- Badges
- Science topic
- Similar topics
- Computer Science
- Images