0

I have a 5120×5120 image . Instead of using for-loop to iterate every pixes,I use matrix to optimise it.However,it turns out that the function remap itself consumes the most of time.So,how can i speed up the remap function?

StoizeMuing
  • 1
  • Are you testing the optimized / release build? I ask because in the past I have seen an enormous slow down in debug mode of Visual Studio. – drescherjm May 11 '22 at 11:22
  • You didn't mentiob you flatform. If it's an Intel based PC, IPP is faster than opencv in some cases. See here: https://www.intel.com/content/www/us/en/develop/documentation/ipp-dev-reference/top/volume-2-image-processing/image-geometry-transforms/geometric-transform-functions/remap.html – wohlstad May 11 '22 at 11:33
  • 1
    needs debugging details. please present a [mre] and timing measurements and all relevant information on the computer you ran this on. -- @wohlstad AFAIK OpenCV can use IPP. both OpenCV and IPP are run by Intel. they'll make sure things run as fast as possible on Intel. – Christoph Rackwitz May 11 '22 at 11:54
  • @ChristophRackwitz, I recall cases in the past when using ipp directly was faster. But it's quite a long time ago, could be that opencv was without ipp then (or at least the version I used was built without it). – wohlstad May 11 '22 at 11:58
  • definitely. it may or may not be built with IPP. though I would expect their OpenCL to run comparably, even if it's not hand-sewn assembly code. – Christoph Rackwitz May 11 '22 at 19:54

0 Answers0