I was wondering:
- How can Intel IPP make OpenCV applications faster?
- How can I integrate it in already written in already written OpenCV applications?
I didn't find an immediate answer (this document doesn't really say much), but finally I found this question which answer has a clear example which answer to question 1. and partially to 2.
However, I have an observation:
- Could the
(const Ipp8u*)&img.data[0]casting considered inefficient? - I want to improve SURF descriptor exploting parallel computing and vectorization. Could IPP help me?
- How can I develop my OpenCV in both cases that IPP is present on the target machine or not?