draw four lines touching the circle, each pair at 90° draw a line splitting each 90° in half (at 45° that is) going through the circle
the intersection of those two midlines should be the centre of the circle :-) better: fit a square around the circle ... Karsten > Gesendet: Dienstag, 05. Januar 2021 um 18:03 Uhr > Von: "Sonali Warunjikar" <sonali.warunji...@gmail.com> > An: debian-med@lists.debian.org > Betreff: Re: Acquiring Dental RVG on Linux > > On Sat, Jan 02, 2021 at 05:24:15PM +0530, Sonali Warunjikar wrote: > > If I imagine line pair to be equivalent of 2 pixels, the pixel width isn't > > matching with that. Essentially lp/mm is appearing coarser than pixel > > width and probably hence called 'true (measured) resolution' and may be > > they aren't expected to match. > > > > I tried both: 1. using pixel width compute pixels per meter 2. using > > lp/mm, assuming 1 lp is equivalent to 2 pixels compute pixels per meter. > > > > The latter is coming closer to the reality, but not sure it's accurate > > enough as yet. May be there was a gap between the finger and the sensor > > when trial x ray was taken as our attention was on usb interface rather > > than lengths (getting computed length smaller than the actual one of the > > portion of the finger). I need to go and gather more x rays carefully (may > > be using a coin to make the boundaries more accurate) before confirming > > which measure is accurate or whether something else is required or is just > > a fixed calibration factor is needed. > > [1] is the image produced by a coin kept right on the sensor whose > diameter is officially 21.93mm (appears 22mm to us by accuracy with which > we can measure at home). > > The spec[2] says: pixel size = 19 micrometer, 'true (measured) resolution > = 16 linepairs/mm. > > The png encoder requires pixels/metre > by pixel size it comes out to be: 52631 > by lines per mm it is (explained above): 32000 > > I measured the count of pixels along diameter line using gimp, by drawing > a line that to a naked eye looks like a diameter. It comes to be 1172. > > So pixels/m should be diameterpixels * 1000/21.93 = 53442.77. This is > closer to the one derived from pixel size, but quite far from linepair/mm. > > Can someone suggest any better way (preferably some tool) to measure the > diameter of the circle in image [1] to improve the accuracy? > > > [1] http://mayuresh.sdfeu.org/coinxray.png > [2] > https://pdf.indiamart.com/impdf/22884448473/MY-16502350/dental-rvg-systems-kodak-5200.pdf > >
