Are not your sampling boxes too much large? Some of them seem to come out from the patches boundaries. Maybe you can try to change the BOX_SHRINK value. Marco 2011/8/5 nino loss <nino@xxxxxxxx>: > I still get a peak error of 18.381637 and an average of 7.896481 ! > > nino > > > On 8/5/2011 3:12 AM, nino loss wrote: >> >> It worked right now with a new image. I made a tiny little bit wider crop >> of the target and corrected -0.16% of rotation. A added -a to be sure. >> >> Now lets see the next step ;-) >> >> nino >> >> >> On 8/5/2011 3:05 AM, nino loss wrote: >>> >>> I already tried without success, because of error code 0x3 "Pattern match >>> wasn't good enough". >>> >>> nino >>> >>> >>> On 8/5/2011 3:03 AM, Graeme Gill wrote: >>>> >>>> nino loss wrote: >>>>> >>>>> I just noticed that the numbers go the wrong way?! >>>> >>>> That would explain it, yes. It's a puzzle why that has happened though. >>>> You might be able to use the scanin -a flag to fix the problem. >>>> >>>> Graeme Gill. >>>> >>> > >