Prev: Substring????
Next: Report Generator IO Documentation
From: dbeg on 2 Mar 2010 15:09
I have RGB images (200*300pixels).
My region of interest ROI in this image is size of 100*100 pixels and its axis can be rotated regarding axis of original image.

My MAIN problem is how to evaluate success of ROI detection (!!).

Let say I have 10 different (!!) images of same object made with scanner at different times.

ROI is detected automatically on all 10 images. Then crop of image is made within ROI -> ROI image.

I manualy select one of sucessfully detected ROI image as reference ROI image.

Now I would like to get the "best fit" of reference ROI image to each of other 9 original images. If there would be no rotation beetwen detected ROIs, normalized 2D correlation between each full image and reference ROI crop would do the yob, but now I have a little problem (big actually).

If there would be no rotation, point of maximum normalized correlation value would give us center of ROI which would serve us for comparision to atomatically detected ROI.

I would really appreciate your advice/suggestion how to get best fit position of rotated crop in original image.

Thank you for help
From: dbeg on 5 Mar 2010 13:39
Was I so unclear, or nobody has come to similar problem?
 | 
Pages: 1
Prev: Substring????
Next: Report Generator IO Documentation