On 11/22/2010 02:24 PM, bioster wrote:
>>> Now, looking through the code I don't see gimp doing anything really
>>> different from me outside>of the convolution matrix code. Frankly, I don't
>>> understand the convolution matrix code.
>>> Why does it take 50 samples and then average them? Is this significant?
>> even if i am not a developer i may assure you that consider the neighboring
>> pixels (that is what convolution matrix are for ) is VERY significant, and
>> sure that may explain well why result look simultaneously more contrasted
>> and smoother.
>> i think you should really give a look to what convolution matrix may do
> Er, I think you misunderstood me a bit. I believe I understand what the
> convolution matrix is, and I also understand why it's important. Yes, the
> entire purpose of the convolution matrix is to look at neighbouring pixels to
> create the pixel you're looking for, but that is not what I was talking about
> when I said '50 samples'.
> In the code which *creates* the convolution matrix it does not look at the
> pixels at all. It creates the convolution matrix using an equation. I would
> think that it would simply plug in the variables for each convolution matrix
> value it's looking for, but what it is actually doing is taking 50 numbers
> and plugs those into the equation, then averages them.
> Just to repeat, this is in the part that creates the convolution matrix, not
> the part that uses the convolution matrix to generate pixel values. I was
> wondering about the purpose of all those samples along the equation that
> generates the matrix.
From the comments in the code, it looks like it's computing the
convolution matrix values by integrating the gaussian bell curve (using
100 points total).
Gimp-developer mailing list