On Fri, Dec 21, 2012 at 9:08 PM, Ulrich Pegelow
<[email protected]> wrote:
> Am 21.12.2012 20:36, schrieb Pascal de Bruijn:
>> Hi,
>>
>> We (Tobias and me) just noticed our sharpen module fails at image
>> borders. This is particularly noticeable (at 200% zoom) with a radius
>> of 8 and amount of 2 and a threshold of 0. In more real-world
>> use-cases it's not so noticeable.
>>
>> Could this be a trivial issue? Or should I create a low priority
>> ticket for the long run?
>>
>
> I think it's expected behavior. For the border pixels we can not apply
> the normal gaussian convolution filter as it would access pixels outside
> the image buffer. For that reason border pixels are just copied from
> input image without sharpening. At extreme parameter settings this leads
> to visible differences.
>
> So, if we would want to change that, we would need to make a special
> round of convolution taking the above mentioned restrictions into
> account, e.g. out of range pixels are replaced by edge values etc.
>
> I doubt if under real life conditions this really makes a difference and
> is worth the effort.

Ok. I would agree with that. So never mind then.

> BTW: our OpenCL path does not suffer from this issue as there is a
> hardware-enabled option to replace out-of-range pixels by edge values.

Ah very nifty.

Regards,
Pascal de Bruijn

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
darktable-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/darktable-devel

Reply via email to