While summarizing the supported kernels of OpenVX, I found that the most of the filter ksize is fixed to 3x3.
Is it because the specified 3x3 ksize is enough to support any image filtering that is occurred in the mobile environment?
Or is it because it is just a not-a-mature specification yet?
If it is the second case, do you guyz have plan to expand the size to 5x5, 7x7, … ?
(as you know, OpenCV supports limitless ksize, and it seems useless overhead to me for the embedded vision).
PS. I’m still looking forward to implement some vision applications using OpenVX header and the opensource implementation. When will it be possible?