Once again, MIPAR (http://MIPAR.us) can be extremely valuable here. Using the reject features tool to reject features that are two pixels in either x, y, or z. You can also reject features whose minor axis is 2 pixels or less.

In MATLAB code, I'd suggest using bwlabel to make a labels map of your objects, using regionprops on those labels to measure the bounding box and minor axis dimensions, and then filter features based on any of those dimensions.

Cheers,

John

Make your mask bigger! Then you can count number of 0 and 1 in your mask and make your correct decision.

However, this can also be done simply by applying Morphological operators, such as: Erosion and Dialation. In these operators selecting a proper structure (mask) will cause the desire output.

Here is the link:

http://www.mathworks.com/help/images/morphology-fundamentals-dilation-and-erosion.html?refresh=true

Best, 

Maybe https://en.wikipedia.org/wiki/Hit-or-miss_transform

Thank You for all your answers help me lot.

Hi,  the image should be processed with mathematical morphology operators that are already implemented in Matlab

Yes, morpholical operators in matLab will do. Check up Region of Interest (RoI) , identify your own region of interest and use the ImageFill function.

In MatLab, I would recommend using imclose with seqaure structuring element.

Read the example at the mathworks:

http://www.mathworks.com/help/images/ref/imclose.html

http://www.mathworks.com/help/images/ref/imclose.html