image region border features
in [Matlab]
|
From: subrajeet Mohapatra on 12 May 2010 11:34 Suppose I have a image with two objects. My job is to find out whether the border of the objects are regular in shape or irregular. I mean to say are there any kind of peaks and valley like structures on the border or not. Suppose we take a coin the border surface is smooth and regular. But if we take say circular blade with teeths then its irregular howto do it. Please suggest
From: Ashish Uthama on 12 May 2010 12:03 On Wed, 12 May 2010 11:34:05 -0400, subrajeet Mohapatra <subrajeets(a)gmail.com> wrote: > Suppose I have a image with two objects. My job is to find out whether > the border of the objects are regular in shape or irregular. I mean to > say are there any kind of peaks and valley like structures on the border > or not. Suppose we take a coin the border surface is smooth and regular. > But if we take say circular blade with teeths then its irregular howto > do it. > > Please suggest One approach might be to use EDGE, followed by bwboundaries/regionprops to obtain the 'borders'. Then, define a 'regularity' metric you want to use. One basic idea: Obtain the difference between successive pixel coordinates along the boundary (perimeter) count the number of instances where this difference is more a threshold if this count/total perimeter pixel is greater than another threshold, mark as 'irregular'. You might get a better response with some sample images and whatever code/logic you have already tried.
From: subrajeet Mohapatra on 19 May 2010 08:25 "Ashish Uthama" <first.last(a)mathworks.com> wrote in message <op.vcln75wia5ziv5(a)uthamaa.dhcp.mathworks.com>... > On Wed, 12 May 2010 11:34:05 -0400, subrajeet Mohapatra > <subrajeets(a)gmail.com> wrote: > > > Suppose I have a image with two objects. My job is to find out whether > > the border of the objects are regular in shape or irregular. I mean to > > say are there any kind of peaks and valley like structures on the border > > or not. Suppose we take a coin the border surface is smooth and regular. > > But if we take say circular blade with teeths then its irregular howto > > do it. > > > > Please suggest > > One approach might be to use EDGE, followed by bwboundaries/regionprops to > obtain the 'borders'. > Then, define a 'regularity' metric you want to use. > One basic idea: > Obtain the difference between successive pixel coordinates along the > boundary (perimeter) > count the number of instances where this difference is more a threshold > if this count/total perimeter pixel is greater than another > threshold, mark as 'irregular'. > > You might get a better response with some sample images and whatever > code/logic you have already tried. Thanks for the reply now I want to measure the surface roughness or smoothness the images are http://drop.io/dhkow3e2082
From: Ashish Uthama on 19 May 2010 09:37 On Wed, 19 May 2010 08:25:21 -0400, subrajeet Mohapatra <subrajeets(a)gmail.com> wrote: What are the 'two objects' in this image? Assuming the circular/eppliptical shape in the centre is your object of interest -what are the structures on the top left and bottom right? I cant think of a way to remove them (maybe fit an ellipse to the edge points and use it as a mask?) -Would you define this as regular or not? Can you tweak your image acquisition to increase the distance of these other structures? Another thought I had for measuring smoothness of a boundary. Fit a line to N consecutive points, find out the distance of the N+1th point to this line. Accumulate this distance over the entire boundary. Fine tuning N and the function that you fit to these points (I said 'line', you could use any other polynomial/function) should give you a decent metric to use.
From: subrajeet Mohapatra on 20 May 2010 00:00
"Ashish Uthama" <first.last(a)mathworks.com> wrote in message <op.vcyf33nga5ziv5(a)uthamaa.dhcp.mathworks.com>... > On Wed, 19 May 2010 08:25:21 -0400, subrajeet Mohapatra > <subrajeets(a)gmail.com> wrote: > > What are the 'two objects' in this image? > Assuming the circular/eppliptical shape in the centre is your object of > interest > -what are the structures on the top left and bottom right? I cant think > of a way to remove them (maybe fit an ellipse to the edge points and use > it as a mask?) > -Would you define this as regular or not? > Can you tweak your image acquisition to increase the distance of these > other structures? > > > Another thought I had for measuring smoothness of a boundary. Fit a line > to N consecutive points, find out the distance of the N+1th point to this > line. Accumulate this distance over the entire boundary. Fine tuning N and > the function that you fit to these points (I said 'line', you could use > any other polynomial/function) should give you a decent metric to use. > Actually this is a complete cell we have to measure the roughness of the entire boundary. The whole violet part is the object of interest. plz check this is it possible with this images. Region of interest is violet part only |