I have acquired several ultrasound images with different pin sizes, for a validation test of mine.
I know the real pin size, and that would be my ground truth. The problem is that when I plot the results of the segmented image (see attached image) in excel file (I export numerical pixel values to excel, the green mask contour, and just plot it), the obvious bottom line where the standoff polymer meets the table and is delined by the green contour is not actually the correct boundary for the accurate pin height measurement.
For instance, if I define the bottom row, where the contour lies horizontally, then I get around 0,42 mm error. The true boundary is approximately 10 pixels above the contour line. How can I justify this to my professor? He won't accept it. As he can see the visible green boundary, at the specific line, he will believe that this line is the correct no matter what and it's me that I must doing something wrong.
But that's not the truth. I have tried hundreds of times and every time I get around that error when I define bottom row as where the contour lies.
I just need to find a reason to justify to my supervisor, that the actual boundary of standoff-desk is approximately 10 pixels higher than it shows and the bright constrast that is being seen on the image is just artifact.
Any other explanation please?