some questions about the library about nvision-cpp HOT 1 OPEN

Hi @yochju,
thanks for your feedback. I'll get to the points individually

Regarding point 1.:
Thanks, I just fixed the examples in the README.md. As you already guessed those examples were outdated relicts. I still have not found a way to link code from the actual example source files (in examples/) into the README.md (so that they stay up to date). The examples in the examples/ directory should all work if your build system is correctly configured (they are built by the CI).

Regarding point 2.:
Your build system should pass the <path-to-repo>/include/ directory as include directory to the compiler, then all paths should be resolved automatically. In terms of CMake you would do something like

include_directories("<path-to-nvision-repo>/include/")

In any case regardless of your build system, there should be no need to adjust any include paths in the source code.
If you need any more assistance on this topic let me know.

Regarding point 3:
Not on on the short term, sorry. I actually wanted to implement some more advanced optical flow algorithms, but I just did not get the time and motivation to do so.
If you feel like, you can implement your linked one and file a Pull Request. Contributions are welcome, just stick to the code style and optionally (in the best case) provide some unit tests and / or examples.

Regarding point 4:
If you want something simple, you can do what all the filter implementations basically do (e.g. when you have a look at the central differences filter class):

• you create the kernel you need (using Eigen), you can set the size (compile time) and coefficients to what you need
• call nvision::image::correlate on your image with that kernel (see here)
• pass a specific BorderHandling functor to correlate if you want something else than "reflect" (see here)

Example (no guarantee that it compiles):

// create kernel with custom size and coefficients
auto kernel = Eigen::Matrix<float, 5, 1>{};
kernel << -1, -1, 0, 1, 1;
// use constant border values with value 1
auto borderHandling = nvision::BorderConstant<float>{1};
// important image type should be explicit otherwise you will store a expression
nvision::Image<nvision::Grayf> resultImage =  nvision::image::correlate(image, kernel, borderHandling);

from nvision-cpp.