This may be related to the new "rounding mode" input, and is a strongly annoying bug. In previous IMAQ version, it was possible to use floating point values to build specific kernels. However, although the kernels are still passed as single precision arrays, things have changed in version 8.6 : non-integer kernel values gives unexpected results. After playing a while, I still cannot understand what Imaq Convolute is doing now. Of course, it it still possible to find workarounds, for instance by using integer multiples of any fractionnal kernel, but re-use of ancient vi's is strongly impaired. Nearly a show stopper... Any comment ?
Edit : I should have post this in the Vision board. I'll post a link to this thread instead...
Message Edité par chilly charly le 08-31-2008 05:43 PM
Chilly Charly (aka CC)
