Area Measurement with Subpixel Accuracy

[Eblanco]

I am planning an area measurement application (high contrast image of a backlit 2D object) and I wonder if it is possible to use somehow subpixel accuracy to improve performance. So far, I have only thought of using 2 steps to calculate the area ( binary thereshold + particle analysis) but I am no sure that this is the best way and there is no subpixel accuracy. Two questions: 1. what about subpixel accuracy for area measurement? is it possible? 2. any suggestion on how to get a better area measurement? Best Regards.

[BruceAmmons]

There is no sub-pixel measurement for area.  Usually you convert it to binary, then just count the number of pixels.

If you know the shape of the area, you could use that to your advantage.  For example, you can find the diameter of a circle or the sides of a rectangle.  This would give you a better measurement.

If it is the only object, you might be able to use the histogram for a sub-pixel type measurement.  Instead of thresholding at a fixed level, you could pick two levels, A and B, which correspond to zero and 100 percent.  All the pixels above B are counted at 100%, and all pixels below A are not counted.  The ones between A and B would be weighted by their intensity such that A=0 and B=1.  This might give you a slightly better value of the area, but probably isn't worth the effort.

Bruce