some help understanding Multi-band level.  

Share your tips and tricks here or get help with any Autopano Pro / Giga problem!
No bug reports (of any kind) in this forum!
no avatar
mgg
Member
 
Topic author
Posts: 28
Likes: 0 post
Liked in: 0 post
Joined: Fri Jun 24, 2011 10:05 pm
Info

some help understanding Multi-band level.

by mgg » Fri Mar 02, 2012 4:18 pm

From the APG documentation:

Multi-band level: Lets you adjust the influence zone for each input pixel. Each pixel will double its influence zone for each supplemental level. For example, at level 2, each pixel acts on a span of 2 pixels in the final rendering, 4 pixels at level 3, 8 pixels at level 4... Level 0 represents the maximal influence zone (depending on the size of the input images). The negative levels let you reduce this area based on the maximum area. -4, the default value, seems to be the best choice in the majority of cases, if the color correction is enabled.

I'm missing something. From experience, it seems that a value of 0 provides the best blending. The second sentence seems to say a larger positive number blends over a larger area. The last seems to say, a negative number decreases the area. The middle sentence says 0 is maximum.... am I the only one that is confused?

My aim here is to get the best blending between images. I used to have terrible issues with vertical banding - which careful vignetting correction in Camera Raw helps a lot. Using "0" for Multi-band helps as well but sometimes I still end up with very subtle bands in some pictures in some places. There is no lens profile in ACR for my camera (Panasonic G3 with either lens), so I'm going to use their tools to create a profile: maybe that will get me better vignetting correction.

So is "0" truly the best value to use for this situation?
tnx
mgg

no avatar
mediavets
Moderator
 
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

by mediavets » Fri Mar 02, 2012 4:38 pm

mgg wrote:There is no lens profile in ACR for my camera (Panasonic G3 with either lens), so I'm going to use their tools to create a profile: maybe that will get me better vignetting correction.

mgg

Does DxO support your camera/lens combinations? If so it may provide good/better vignetting correction.

http://www.dxo.com/uk/photo
Last edited by mediavets on Fri Mar 02, 2012 4:38 pm, edited 1 time in total.
Andrew Stephens
Many different Nodal Ninja and Agnos pano heads. Merlin/Panogear mount with Papywizard on Nokia Internet tablets.
Nikon D5100 and D40, Sigma 8mm f3.5 FE, Nikon 10.5mm FE, 35mm, 50mm, 18-55mm, 70-210mm. Promote control.

no avatar
mgg
Member
 
Topic author
Posts: 28
Likes: 0 post
Liked in: 0 post
Joined: Fri Jun 24, 2011 10:05 pm
Info

by mgg » Fri Mar 02, 2012 4:46 pm

The G3 and 14-42 lens is supported, but the 45-200 lens isn't even listed as a future.......

tnx
mgg

no avatar
mgg
Member
 
Topic author
Posts: 28
Likes: 0 post
Liked in: 0 post
Joined: Fri Jun 24, 2011 10:05 pm
Info

by mgg » Tue Mar 06, 2012 9:28 pm

So, no one truly understands that paragraph? Or everyone understands it except me?

"Multi-band level: Lets you adjust the influence zone for each input pixel. Each pixel will double its influence zone for each supplemental level. For example, at level 2, each pixel acts on a span of 2 pixels in the final rendering, 4 pixels at level 3, 8 pixels at level 4... Level 0 represents the maximal influence zone (depending on the size of the input images). The negative levels let you reduce this area based on the maximum area. -4, the default value, seems to be the best choice in the majority of cases, if the color correction is enabled."

User avatar
DrSlony
Moderator
 
Posts: 1893
Likes: 0 post
Liked in: 0 post
Joined: Sat Nov 03, 2007 6:30 pm
Location: Sweden
Info

by DrSlony » Mon Mar 12, 2012 9:40 pm

I too would like a clarification of "The negative levels let you reduce this area based on the maximum area."
In the case of a 4000x3000 photo,
...
-4 ? (if 0 is 2048, then this is 128?)
-3 ? (if 0 is 2048, then this is 256?)
-2 ? (if 0 is 2048, then this is 512?)
-1 ? (if 0 is 2048, then this is 1024?)
0 max (2048? "Level 0 represents the maximal influence zone (depending on the size of the input images)")
1 1
2 2
3 4
4 8
5 16
6 32
7 64
8 128
9 256
10 512
11 1024
12 2048
13 4096 (larger than image, so doesn't exist)

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Tue Mar 13, 2012 9:15 am

Really good explaination DrSlony. Thanks.
One note : if you have APG, you can go to exposure fusion dialog box where you can see in live the influence of this value ( multiband level slider at the bottom ).

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Tue Mar 13, 2012 2:32 pm

mgg wrote:.... am I the only one that is confused?

No. Definitely not . . . :cool: I also can only guess what´s meant here in detail and what setting does which influence on our working.

best, Klaus
Simplicity is the keynote of all true elegance. Coco Chanel

User avatar
DrSlony
Moderator
 
Posts: 1893
Likes: 0 post
Liked in: 0 post
Joined: Sat Nov 03, 2007 6:30 pm
Location: Sweden
Info

by DrSlony » Wed Mar 14, 2012 10:40 pm

AlexandreJ: explanation? I was asking :D

The exposure fusion dialog only allows one to use negative numbers: -10 to 0.

If you confirm what I wrote, it would mean that the negative range actually does the same thing as the positive range, just 'from the other end', so no problem really.

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Thu Mar 15, 2012 4:05 pm

A pyramid has a level ( height ). This level ( height ) depends on input image size. For a 4000x3000 photos, the level is around 12 ( or 13 depending on the rounding ) : let's say 12.
+X means : X levels from the bottom : Example : 5 means 5 levels, so the blending will occurs over 16 pixes around edges.
-Y means : Y levels from the top. Example -2 means do all levels except last 2. In this case, 10, so everything is smooth at 512 pixels at least.

Of course, both positives values and negative values are the same. For a 12 levels pyramid : 5 = -7, 8 = -4, etc.
0 is an exception : it means all levels.

User avatar
gkaefer
Member
 
Posts: 3549
Likes: 0 post
Liked in: 15 posts
Joined: Tue Jun 09, 2009 1:01 pm
Location: Salzburg
Info

by gkaefer » Thu Mar 15, 2012 5:51 pm

AlexandreJ wrote:A pyramid has a level ( height ). This level ( height ) depends on input image size. For a 4000x3000 photos, the level is around 12 ( or 13 depending on the rounding ) : let's say 12.
+X means : X levels from the bottom : Example : 5 means 5 levels, so the blending will occurs over 16 pixes around edges.
-Y means : Y levels from the top. Example -2 means do all levels except last 2. In this case, 10, so everything is smooth at 512 pixels at least.

Of course, both positives values and negative values are the same. For a 12 levels pyramid : 5 = -7, 8 = -4, etc.
0 is an exception : it means all levels.

hm. ok now I should say I'm wiser...
... but sorry cant follow the calculation

ok asume a height of 12 and a value of +X = 5 ... how do you get to 16pixels. and how to use this If I asume +X =6 ???
Georg

no avatar
lumelix
Member
 
Posts: 528
Likes: 0 post
Liked in: 0 post
Joined: Mon Oct 25, 2010 5:51 pm
Location: Switzerland
Info

by lumelix » Thu Mar 15, 2012 7:39 pm

Alexandre
A picture says more than thousand words :)
I think that kolor should illustrate this tricky thing with the pyramid graphically, so that we can understand this definitely.
I have seen something like this in the wiki earlier, but I couldn't find it again.
Last edited by lumelix on Thu Mar 15, 2012 10:36 pm, edited 1 time in total.
Regards
Martin

User avatar
DrSlony
Moderator
 
Posts: 1893
Likes: 0 post
Liked in: 0 post
Joined: Sat Nov 03, 2007 6:30 pm
Location: Sweden
Info

by DrSlony » Sun Mar 18, 2012 6:39 pm

Alexandre you explained it very well now, thank you.

gkaefer the 5th level covers 16 pixels. See my post #5

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Mon Mar 19, 2012 9:01 am

BTW : we are working on the documentation to make this part full and better understandable :
WIP here : http://www.autopano.net/wiki-en/action/view/Understanding_and_using_the_rendering_engine

User avatar
gkaefer
Member
 
Posts: 3549
Likes: 0 post
Liked in: 15 posts
Joined: Tue Jun 09, 2009 1:01 pm
Location: Salzburg
Info

by gkaefer » Mon Mar 19, 2012 4:13 pm

AlexandreJ wrote:BTW : we are working on the documentation to make this part full and better understandable :
WIP here : http://www.autopano.net/wiki-en/action/view/Understanding_and_using_the_rendering_engine

Hi Alexandre,

looking at the new docu of your link and in special at the image

http://www.autopano.net/wiki-en/action/view/Image:Schema_Muliband_level_en.png

I've following question:

the size in poxel of the influence zone is the common denominator between the + and the - values.

why not using a scale from 1 to 12 (or 0 to 11) where the smallest is representing the 1 pixel and the highest the 4096 pixel?

with current scaling I have 21 possible values I can set but 9 of them can be set with 2 different values (the -10 and the +2 is representing the 2 pixel, the -9 and the +3 is representing the 4pixel etc. and I assume using +2 or the -10 will give the same result, because both are using the influence zone of 2 pixels?

Liebe GrüàŸe,
Georg
Last edited by gkaefer on Mon Mar 19, 2012 4:14 pm, edited 1 time in total.

User avatar
gkaefer
Member
 
Posts: 3549
Likes: 0 post
Liked in: 15 posts
Joined: Tue Jun 09, 2009 1:01 pm
Location: Salzburg
Info

by gkaefer » Mon Mar 19, 2012 4:16 pm

DrSlony wrote:Alexandre you explained it very well now, thank you.

gkaefer the 5th level covers 16 pixels. See my post #5

thanks - got it ;-)

User avatar
klausesser
Member
 
Posts: 8836
Likes: 5 posts
Liked in: 64 posts
Joined: Mon May 22, 2006 12:18 am
Location: Duesseldorf, Germany
Info

by klausesser » Mon Mar 19, 2012 6:10 pm

AlexandreJ wrote:BTW : we are working on the documentation to make this part full and better understandable :
WIP here : http://www.autopano.net/wiki-en/action/view/Understanding_and_using_the_rendering_engine

Well explained! Thank you. As i understand level "0" is the best to use in my case with the sky images? So Vincèn was right?
When i use it the borders seem become a bit sharper than with -4 (default).

best, Klaus

The best result i got is this one - using -4 but some anchors:


Simplicity is the keynote of all true elegance. Coco Chanel

User avatar
[bo]
Member
 
Posts: 1226
Likes: 0 post
Liked in: 0 post
Joined: Fri May 05, 2006 8:16 am
Location: Bulgaria
Info

by [bo] » Wed Mar 21, 2012 11:12 am

OT: I've never seen more icons on the dock or desktop! Amazing! How do you work and find things :)
Some of my panoramas, posted in the Autopano Pro flickr group.

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Wed Mar 21, 2012 2:27 pm

gkaefer wrote:the size in poxel of the influence zone is the common denominator between the + and the - values.

why not using a scale from 1 to 12 (or 0 to 11) where the smallest is representing the 1 pixel and the highest the 4096 pixel?

with current scaling I have 21 possible values I can set but 9 of them can be set with 2 different values (the -10 and the +2 is representing the 2 pixel, the -9 and the +3 is representing the 4pixel etc. and I assume using +2 or the -10 will give the same result, because both are using the influence zone of 2 pixels?

The pyramid has a variable max level ( sometimes, it can be only 8, other time, it can grow up to 12 or 13 ). So the total number of value can be larger of smaller than 21.
Depending on that, the association ( -something = +some other value ) changes. We want it simple without anyone knowing that there is something not constant around.
That's why we came up with a (-10 | 0 | +10 ) slider. We though also about 3 checks box ( small, normal, large uniformisation ), but that didn't give the accuracy I wanted.

no avatar
marzipano
Member
 
Posts: 431
Likes: 1 post
Liked in: 14 posts
Joined: Sat Mar 05, 2011 1:32 pm
Location: Richmond London UK
Info

by marzipano » Sat Mar 24, 2012 10:41 am

AlexandreJ wrote:
gkaefer wrote:the size in poxel of the influence zone is the common denominator between the + and the - values.

why not using a scale from 1 to 12 (or 0 to 11) where the smallest is representing the 1 pixel and the highest the 4096 pixel?

with current scaling I have 21 possible values I can set but 9 of them can be set with 2 different values (the -10 and the +2 is representing the 2 pixel, the -9 and the +3 is representing the 4pixel etc. and I assume using +2 or the -10 will give the same result, because both are using the influence zone of 2 pixels?

The pyramid has a variable max level ( sometimes, it can be only 8, other time, it can grow up to 12 or 13 ). So the total number of value can be larger of smaller than 21.
Depending on that, the association ( -something = +some other value ) changes. We want it simple without anyone knowing that there is something not constant around.
That's why we came up with a (-10 | 0 | +10 ) slider. We though also about 3 checks box ( small, normal, large uniformisation ), but that didn't give the accuracy I wanted.

I understand how the current system is meant to work now but I would say it suffers from an unnecessary over-complication in using values both above and below zero. I cannot see why the slider cannot just go from 0 to 12 where 0 represents max blending equal to the pixel width of the image itself. Then 1=1/2 max 2=1/4 max 3=1/8 max and so on up to 12 or 13 as required. The actual pixel values themselves would not be that important but for a 4096 pixel width it would be 4096=0, 2048=1, 1024=2, 512=3, 256=4, 128=5, 64=6, 32=7, 16=8, 8=9, 4=10, 2=11, 1=12, 1=13

The current system is also not intuitive because in moving the slider from -1 to 0 to 1 (which are all next to each other on the scale) you move through a discontinuity for example where 0=4096 it goes 2048 to 4096 to 1. The slider goes from max to min in one jump there

regards,
Martin

User avatar
gkaefer
Member
 
Posts: 3549
Likes: 0 post
Liked in: 15 posts
Joined: Tue Jun 09, 2009 1:01 pm
Location: Salzburg
Info

by gkaefer » Sat Mar 24, 2012 11:11 am

AlexandreJ wrote:
gkaefer wrote:the size in poxel of the influence zone is the common denominator between the + and the - values.

why not using a scale from 1 to 12 (or 0 to 11) where the smallest is representing the 1 pixel and the highest the 4096 pixel?

with current scaling I have 21 possible values I can set but 9 of them can be set with 2 different values (the -10 and the +2 is representing the 2 pixel, the -9 and the +3 is representing the 4pixel etc. and I assume using +2 or the -10 will give the same result, because both are using the influence zone of 2 pixels?

The pyramid has a variable max level ( sometimes, it can be only 8, other time, it can grow up to 12 or 13 ). So the total number of value can be larger of smaller than 21.
Depending on that, the association ( -something = +some other value ) changes. We want it simple without anyone knowing that there is something not constant around.
That's why we came up with a (-10 | 0 | +10 ) slider. We though also about 3 checks box ( small, normal, large uniformisation ), but that didn't give the accuracy I wanted.

ok, got it. my impression is that the user currently does not know which size the currently used images does have, 4096, 2048 etc. and so user cant predict where the best results can be expected rather with minus rather with plus values. so its still a bit inefficient a try & error playing...

so a colorscale from green to red obove/down the (current) slider could help the user where the best results can be expected. the scale should be dynamically & dependant to the internally detected pixel size?

Liebe Gruesse,
Georg

User avatar
AlexandreJ
Kolor Team
 
Posts: 5987
Likes: 7 posts
Liked in: 10 posts
Joined: Mon Nov 14, 2005 4:56 pm
Location: Francin, France
Info

by AlexandreJ » Mon Mar 26, 2012 8:14 am

Yes Gkaefer, perhaps, that's another option. We did this single slider for compacity goal. I agree with the discontinuity.
If you have another UI to propose, I didn't find anything better yet.

no avatar
marzipano
Member
 
Posts: 431
Likes: 1 post
Liked in: 14 posts
Joined: Sat Mar 05, 2011 1:32 pm
Location: Richmond London UK
Info

by marzipano » Tue Mar 27, 2012 11:20 am

AlexandreJ wrote:Yes Gkaefer, perhaps, that's another option. We did this single slider for compacity goal. I agree with the discontinuity.
If you have another UI to propose, I didn't find anything better yet.

As I proposed in my post #19 the use of 2 scales is an over-specification of the problem IMO. One scale with a slider going from 0 to +n is all that is required with each increment reducing the scope of the smoothing by an exponential factor of (1/2) to the power n

This is also the best solution in terms of "compacity" (compactness)


Who is online

Users browsing this forum: No registered users and 2 guests