On Thu, Nov 22, 2012 at 12:20 PM, Petr Dlouhý <petr.dlo...@email.cz> wrote: > Hi,
Hi Petr, > the image seems to be much better put together and I have obviously better > match rate. Although sometimes even clearly scanned fingerprints fail to be > matched. I think, that the image is bit too light and slightly blurry, It's blurry due to bi-linear filtration. Image is stretched 2x in horizontal dimension to make libfprint algorithms work better on it. >which leads to false gaps in the lines. Sometimes, some of the stripes are a >bit darker or lighter than the rest of the image. Could you send me (in private mail) screenshot of fprint_demo (with black square in the middle of scan to prevent fingerprint data disclosure would be OK, see screenshot from http://anarsoul.blogspot.com/2012/10/fingerprinting-again.html for example), so I get idea what are you talking about. Both normal and binarized versions would be nice. Btw, sensor area is rather small (128px only), so I suspect it's all we can get from this device. However I'm getting pretty good results with it (up to 30-35 minutiaes) and score for match is usually higher than 50, it's about 30 if scan is not accurate enough, but still enough to consider if finger matches or not. Regards Vasily _______________________________________________ fprint mailing list fprint@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/fprint
