Recoding openCV C++ project in pure verilog

[promach]

I have a 3 month project to work with by hand-recoding openCV C++ project into pure RTL for FPGA usage.

I have a Xilinx Zynq FPGA and I have Vivado.

The code I am using are at https://github.com/Itseez/opencv/blob/master/modules/features2d/src/orb.cpp , https://github.com/Itseez/opencv/blob/master/modules/features2d/src/feature2d.cpp , and http://pastebin.com/N69WE89j  The full eclipse C++ project can be found in the attachment as well.

Anyone can advise on how to start working on this ? I have been told to scrutinise the code for any maths operations which I am doing now and I need to use the AXI interface and logicore IP for this, right ?

Thanks !

[promach]

So, the focus here is to implement floating-point maths operation in FPGA ? Anyone can recommend good references for this ?

I searched online and found that equation (1) can be transformed into (2). But how ?

Pixel_ C = (int)(Pixel_A * alpha + Pixel_B * ( 1-alpha ));                ---> (1)
Pixel_ C = (Pixel_A * alpha + Pixel_B * ( 32-alpha ) + 16)>>5;     ---> (2)

[hate]

I believe translating OpenCV code into RTL directly won't be any efficient. You need to implement your algorithms suitable to and taking advantage of an FPGA architecture. But maybe you need to do this for study purposes?

I searched online and found that equation (1) can be transformed into (2). But how ?

Pixel_ C = (int)(Pixel_A * alpha + Pixel_B * ( 1-alpha ));                ---> (1)
Pixel_ C = (Pixel_A * alpha + Pixel_B * ( 32-alpha ) + 16)>>5;     ---> (2)

This may be related to how OpenCV stores alpha data.