[Bf-cycles] Motion blur related typo?

[Nate Wiebe]

Sorry for so much mail, but here are the tests summarized in image form.
Really, the patch makes MB usable. Currently, I'd prefer Full Sampled
Motionblur over what's in trunk.

Test1:

SVN:
[image: Inline image 1]

Patch:
[image: Inline image 2]

Test2:

SVN:
[image: Inline image 3]

Patch:
[image: Inline image 4]


Test3:

SVN:
[image: Inline image 5]

Patch:
[image: Inline image 6]


The only time that trunk performed properly was in Test 2. The other two
times it was completely wrong.

This seems like an obvious decision to me, but I may be missing something.
Ultimately you should pick the solution that is the most correct since
currently we don't have a perfect solution. Please let me know of corner
cases where the patch would cause problems so I can update my regression
file.

-NateW


On Sun, Nov 11, 2012 at 12:24 AM, Nate Wiebe <natewiebe13 at gmail.com> wrote:

> Here is a quick blend I set up with 3 tests. They are in different scenes.
> With the patch, they all perform like I would expect.
>
> Blend: http://www.pasteall.org/blend/17424
>
> -NateW
>
>
>
> On Sun, Nov 11, 2012 at 12:15 AM, Nate Wiebe <natewiebe13 at gmail.com>wrote:
>
>> storms change does fix that weird blur problem when rotating around an
>> object (bug #32974). I did a couple tests and it did seem to give better
>> results. Are there any cases where the proposed patch gives bad results?
>>
>>
>>
>> On Sat, Nov 10, 2012 at 11:58 PM, storm <kartochka22 at yandex.ru> wrote:
>>
>>> I get it, but my little scene looks more correct with that "fix", maybe
>>> because changing only w in 4D projective space will make strange big
>>> blur along axis ? Maybe do not touch "decompose.y" at all?
>>>
>>> В Вс., 11/11/2012 в 04:07 +0100, Brecht Van Lommel пишет:
>>> > quick arc-length parametrization of the curve
>>>
>>>
>>> http://math.stackexchange.com/questions/12186/arc-length-of-bezier-curves
>>>
>>> Really hard task, elliptic integrals, something that i prefer to skip :P
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20121111/73009595/attachment.htm