Talk:Roberts cross

Untitled
It's a good idea to provide computational steps or a link on how to compute these basic operators since the article assumes that the reader knows how to apply a kernel.

Is this right?
I'm not incredibly familiar with the Roberts Cross Operator but reading through the article,

Shouldn't

$$ z_{i,j} = \sqrt{(y_{i,j} - y_{i+2,i+1})^2 + (y_{i+1,j} - y_{i, j+1})^2 } $$

be

$$ z_{i,j} = \sqrt{(y_{i,j} - y_{i+1,j+1})^2 + (y_{i+1,j} - y_{i, j+1})^2 } $$

??? —Preceding unsigned comment added by 198.151.13.10 (talk) 18:13, 13 August 2010 (UTC)


 * OK, now who's gonna fix it? --Volyrkr (talk) 17:26, 19 August 2010 (UTC)

One more thing that I noticed: why the first square root is applied?

$$ y_{i,j} = \sqrt{x_{i,j}} $$

Shouldn't it just be:

$$ z_{i,j} = \sqrt{(x_{i,j} - x_{i+1,j+1})^2 + (x_{i+1,j} - x_{i, j+1})^2 } $$

I just read the original paper of Lawrence Roberts and there's no mention whatsoever for a first square root step. It is just kernel convolution and then gradient calculation using the formula above.

If anyone has any idea on the matter, please let me know. — Preceding unsigned comment added by 2A02:587:A569:2100:5DD3:580A:58A8:2040 (talk) 16:01, 26 October 2020 (UTC)

Costella operator
I removed references to the Costella operator per discussion at Talk:Sobel_operator. — Control.valve (talk) 18:34, 12 September 2011 (UTC)