[Bf-cycles] Question about Seed

[Brecht Van Lommel]

The reason they might be different is that we're using a quasi monte
carlo sequence, which is better distributed than purely random
numbers. If these were actually random numbers, it would not matter.
That being said, I'm not sure how much difference it makes in practice
when combining images with thousands of samples.

Brecht.

On Wed, Nov 23, 2011 at 7:52 AM, Julian Hhhhhh
<julian.hhhhhh at googlemail.com> wrote:
> Thomas: Could you render a 3000 sample Version and compare some pixel-values
> by hand? I still believe the result should be identical (unless the renderer
> already shoots more photons in noisier areas - don't know it that's already
> supported (but you probably do).
>
> Am 22.11.2011 23:34 schrieb "Thomas Dinges" <blender at dingto.org>:
>>
>> Original 1000s http://temp.dingto.org/render_seed/1000.png
>> Original 2000s http://temp.dingto.org/render_seed/2000.png
>>
>> Combined 3000s, with a Weighting of 0.66:
>> http://temp.dingto.org/render_seed/3000_combined.png
>>
>> Am 22.11.2011 23:20, schrieb Jordan Miller:
>>
>> well the way to test would be to do the 3600 sample render and compare the
>> results. you can do a convert to grayscale then Boolean subtract the two
>> images, then invert the result to see pixel differences.
>> I suspect the 3600 sample image will be superior. should be identical
>> render time but require no post processing.
>> now I want to see the image in question, too! care to post it?
>> jordan
>>
>>
>>
>> On Nov 22, 2011, at 5:15 PM, Thomas Dinges <blender at dingto.org> wrote:
>>
>> Yes, that's what Jordan suggested. :)
>> This makes sense and also gives better results.
>>
>> I hope we will get a better solution though, a good merging system would
>> be awesome. :)
>>
>> Am 22.11.2011 23:11, schrieb Kel M:
>>
>> On Tue, Nov 22, 2011 at 4:59 PM, Thomas Dinges <blender at dingto.org> wrote:
>>>
>>> That seems to give better results, right. Still not sure about this.
>>>
>>> Am 22.11.2011 22:36, schrieb Jordan Miller:
>>> > have you tried merging them weighted ("factor" setting) proportionally
>>> > to the number of samples?
>>> >
>>> > jordan
>>> >
>>> >
>>> >
>>> >
>>> >
>>> > On Nov 22, 2011, at 4:19 PM, Thomas Dinges wrote:
>>> >
>>> >> Hi Brecht,
>>> >> I am testing the Seed function atm in the Integrator panel.
>>> >>
>>> >> Am I right that images, rendered with different seeds can be merged
>>> >> together?
>>> >> So for example:
>>> >> Image_1 has Seed 1 and 1000 Samples
>>> >> Image_2 has Seed 5 and 2000 Samples
>>> >>
>>> >> Do I have 3000 Samples after merging them together?
>>> >> I did that in the compositor with the Mix node, factor 0.5 and I used
>>> >> OpenEXR renders for it.
>>
>> If the second image was in the second input, shouldn't you set the Factor
>> to 0.66? :)
>>
>>>
>>> >> The combined result was better than the single images, I have the
>>> >> impression (more noise free).
>>> >>
>>> >> Then I rendered another image with 600 samples, and combined it to the
>>> >> already combined result. Now I have the impression though that the
>>> >> noise
>>> >> became more again, rather than less.
>>> >>
>>> >> So do I something wrong, or is that not supposed to work the way I use
>>> >> it? :)
>>> >>
>>> >> Best regards,
>>> >> Thomas
>>> >>
>>> >> --
>>> >> Thomas Dinges
>>> >> Blender Developer, Artist and Musician
>>> >>
>>> >> www.dingto.org
>>> >>
>>> >> _______________________________________________
>>> >> Bf-cycles mailing list
>>> >> Bf-cycles at blender.org
>>> >> http://lists.blender.org/mailman/listinfo/bf-cycles
>>> > _______________________________________________
>>> > Bf-cycles mailing list
>>> > Bf-cycles at blender.org
>>> > http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>> --
>>> Thomas Dinges
>>> Blender Developer, Artist and Musician
>>>
>>> www.dingto.org
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>> --
>> Thomas Dinges
>> Blender Developer, Artist and Musician
>>
>> www.dingto.org
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>> --
>> Thomas Dinges
>> Blender Developer, Artist and Musician
>>
>> www.dingto.org
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>