On Mon, Feb 14, 2011 at 9:21 AM, Alex Gomez <[email protected]> wrote:
> Thanks Andrey. As a follow up, do you have a timing description of the > behavior of the memory controller and the compressor? Or is my best best to > run the simulation and look at the behavior there? The distribution should include files for functional simulation with Icarus Verilog and GTK Wave. I do not use timing simulation as even functional is rather slow. For timing I just use attic timing analysis. > > Also, I was able to implement 1dimensional edge detection by modifying > mcontr353.v. > Alex, I would suggest not to additionally complicate memory controller, but rather put processing before it > Now I need to run 2dimensional edge detection, which requires a window of > at least 3x3 pixels. My idea is to use the currently unused channel 1 to > buffer a portion of the image (similar to the compressor buffer), run the > edge detection on 3x3 windows read from the buffer, and then output directly > to memory using channel 3 or a similar scheme (the alternative would be to > write to memory the way the compressor does, but I think that is more > complicated). Then I could check the result using the RAW image download. > What do you think of this plan? Does it seem feasible? > I would recommend using the data in channel 2 (going to compressor) - data there is formatted in 20x20 (overlapping) tiles, so you can use even 5x5, not just 3x3 kernels Andrey > > Thanks again for your help! > > -Alex > >
_______________________________________________ Support-list mailing list [email protected] http://support.elphel.com/mailman/listinfo/support-list_support.elphel.com
