1 | initial version |
The only thing you should do, is to use Android OpenCV Manager. On a Tegra 3 device it will provide your application with optimized version of OpenCV. These are dynamic libraries with API identical to the default version of OpenCV, but with a set of optimizations enabled. Corresponding packages are already published to Google Play market, so you can install your application on a Tegra 3 device and see if it gets performance improvement.
The list of optimized functions is available here: http://docs.opencv.org/opencv2refman-tegra.pdf, check Chapter 17, "TEGRA-POWERED COMPUTER VISION". Please note that not 100% of API was optimized, so probably your bottleneck will not disappear... In such a case you can submit a feature request on the http://code.opencv.org.
And I suggest you to wait for the final OpenCV 2.4.3 release, because it suggests much better experience for end users. I mean significantly reduced number of clicks.
2 | No.2 Revision |
The only thing you should do, is to use Android OpenCV Manager. On a Tegra 3 device it will provide your application with optimized version of OpenCV. These are dynamic libraries with API identical to the default version of OpenCV, but with a set of optimizations enabled. Corresponding packages are already published to Google Play market, so you can install your application on a Tegra 3 device and see if it gets performance improvement.
The list of optimized functions is available here: http://docs.opencv.org/opencv2refman-tegra.pdf, check Chapter 17, "TEGRA-POWERED COMPUTER VISION". Please note that not 100% of API was optimized, so probably your bottleneck will not disappear... In such a case you can submit a feature request on the http://code.opencv.org.
OpenCV for Tegra is also included into the NVIDIA's Tegra Android Development Pack.
And I suggest you to wait for the final OpenCV 2.4.3 release, because it suggests much better experience for end users. I mean significantly reduced number of clicks.