Intel® Integrated Performance Primitives
Deliberate problems developing high-performance vision, signal, security, and storage applications.
Announcements
FPGA community forums and blogs on community.intel.com are migrating to the new Altera Community and are read-only. For urgent support needs during this transition, please visit the FPGA Design Resources page or contact an Altera Authorized Distributor.

cvFilter2D function fails when using IPP

mprobst
Beginner
689 Views
Hi all,

I am using openCV's cvFilter2D function with different filter kernels. When I have IPP installed the resulting images get covered with white/black (depending on the background) stripes (~6 with 8-bit and >10 with 32-bit images). As soon as I remove IPP the function works just fine. Does anybody have an idea what that could be. Could this be a bug in IPP?

Thanks a lot,

Martin.
0 Kudos
2 Replies
Vladimir_Dudnik
Employee
689 Views

Hi Martin,

there is answer from our expert:

As the stripe size depends on the datatype, it looks like a bug in OpenCV wrapping code. The suggestion is to try the latest OpenCV version from CVS. And/or provide the parameters used (filter size, anchor point coordinates, filter coefficients, image size and type (bit-depth and number of channels)).

Regards,
Vladimir

0 Kudos
mprobst
Beginner
689 Views
Hi Vladimir,

thanks very much for your reply. I am already using the newest openCV version (0.9.7 under Linux). OpenCV's cvFilter2D function prototype looks like this:

void cvFilter2D( const CvArr* src, CvArr* dst,
const CvMat* kernel,
CvPoint anchor=cvPoint(-1,-1)),

so the only parameters I can play with are the anchor point. src and dst are images with the same size, number of channels and bit-depth and the kernel I am using is either a 3x3 or 5x5 matrix. I just don't see why it works fine without IPP. Please let me know if you have more ideas. I will try to go through openCV's wrapping code upon time.

Cheers,

Martin.
0 Kudos
Reply