[APG 3.x Windows XP/32] APG 3.x and PTGui Pro ....  

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
mediavets
Moderator
 
Topic author
Posts: 16415
Likes: 2 posts
Liked in: 130 posts
Joined: Wed Nov 14, 2007 2:12 pm
Location: Isleham, Cambridgeshire, UK.
Info

[APG 3.x Windows XP/32] APG 3.x and PTGui Pro ....

by mediavets » Mon Aug 05, 2013 10:50 am

the screenshots below show comparable views of panos - displayed using the DeValVR viewer - stitched from the same image set which comprised two rows of six 9-bracketed exposures fused with Photomatic shot with D800 and 10.5mm FE on a VRDrive2 at +/- 45 degrees pitch.

Stitching done with APG 3.07 and PTGui Pro 9.1.7 using default settings for rendering, each identifiable from their respective watermarks.

Screenshot 1 - nadir area:

I much prefer the way PTGui has chosen to blend the nadir area to the way APG 3.x has done it.

Why is the APG 3.x result so 'bad'?

Sceenshot 2 - desk area:

Interesting to observe that the two programs have chosen to blend the area of the flat screen monitor quite dirreferntly.

Anyone know why?
.............

A further observation, the stitch from PTGui Pro was perfect, the stitch from APG 3.07 was far from perfect.




Last edited by mediavets on Mon Aug 05, 2013 10:52 am, 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
lumelix
Member
 
Posts: 528
Likes: 0 post
Liked in: 0 post
Joined: Mon Oct 25, 2010 5:51 pm
Location: Switzerland
Info

by lumelix » Mon Aug 05, 2013 11:12 am

Hi
I think the colors are much better in the APG-example.
Stitching around the nadir with it's moving panohead-parts is something fatefully and differs from case to case.

Why didn't you use the 16mm fullsize (FX) FE on the D800? It's a really excelent lens.
To improve the stitching accuracy I would take only 5 of the 9 brackets with 2EV steps.
Regards
Martin

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

by mediavets » Mon Aug 05, 2013 11:23 am

lumelix wrote:Hi
I think the colors are much better in the APG-example.

I tweaked the gamma somewhat randomly on both panos to lighten up the nadir area so it could be seen better - so I shouldn't make too much of the colour difference.

Stitching around the nadir with it's moving panohead-parts is something fatefully and differs from case to case.

The difference is consistent between APG 3.x and PTGui pro.

Why didn't you use the 16mm fullsize (FX) FE on the D800? It's a really excelent lens.

They are not my images. the person who shot the iages also has a D90 and initially bought the 10.5 FE for use with that.

To improve the stitching accuracy I would take only 5 of the 9 brackets with 2EV steps.

Why would that improve stitching accuracy?

The bracketed exposures were fused using Photomatic prior to stitching.

Why was the PTGui pro stitch perfect and APG 3.x stitch far from perfect?
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
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Mon Aug 05, 2013 12:35 pm

Same problems at this side of the English channel.....observed today.....the why is easy to answer by the way. Because both programs use a different blending algorithm.....but what's real funny.....take the middle tier of the bracketed set.....if the creator (lets call her Seitzy), has done her work well (no reason for not to), the middle shot of the bracketed row and the exposure fused by Photomatix should like two peas in a pot ((c) Forrest Gump) except for the contrasts in them......now my prediction is that:

1) Middle tear result will be comparable to the PTGui result nadirwise....

2) Middle tear result will be incomparable to the PTGui result HDR wise....

BTW, welcome to the club Andrew...
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

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 Aug 05, 2013 12:51 pm

Artisan New wrote:Because both programs use a different blending algorithm.....

Both apps us different detecting algos AND different blending algos.
That´s why they´re hardly comparable.

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

no avatar
Destiny
Moderator
 
Posts: 7886
Likes: 6 posts
Liked in: 228 posts
Joined: Mon Apr 26, 2010 11:55 pm
Info

by Destiny » Mon Aug 05, 2013 1:05 pm

No.. and the results are hardly comparable either.....

Destiny...

klausesser wrote:
Artisan New wrote:Because both programs use a different blending algorithm.....

Both apps us different detecting algos AND different blending algos.
That´s why they´re hardly comparable.

Klaus

no avatar
Destiny
Moderator
 
Posts: 7886
Likes: 6 posts
Liked in: 228 posts
Joined: Mon Apr 26, 2010 11:55 pm
Info

by Destiny » Mon Aug 05, 2013 1:10 pm

Well Ed.. From now on.. me, myself, and i, not forgetting the very talented Seitzy, agree with everything we say.. :D

Destiny..

Artisan New wrote:Same problems at this side of the English channel.....observed today.....the why is easy to answer by the way. Because both programs use a different blending algorithm.....but what's real funny.....take the middle tier of the bracketed set.....if the creator (lets call her Seitzy), has done her work well (no reason for not to), the middle shot of the bracketed row and the exposure fused by Photomatix should like two peas in a pot ((c) Forrest Gump) except for the contrasts in them......now my prediction is that:

1) Middle tear result will be comparable to the PTGui result nadirwise....

2) Middle tear result will be incomparable to the PTGui result HDR wise....

BTW, welcome to the club Andrew...

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

by mediavets » Mon Aug 05, 2013 1:25 pm

klausesser wrote:
Artisan New wrote:Because both programs use a different blending algorithm.....

Both apps us different detecting algos AND different blending algos.
That´s why they´re hardly comparable.

Klaus

Indeed hardly comparable ...

One (PTGui Pro) produces an apparently perfect result with default settings, and I've hardly used the program ever before.

The other (AutoPG 3.x) doesn't; and it's diifficult if not impossible to get a comparable result even after a lot of manual intervention, which I find disappointing because I like APP/APG and I've been using APP/APG for years.
Last edited by mediavets on Mon Aug 05, 2013 2:16 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.

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 Aug 05, 2013 3:31 pm

For bracketed series, it depends of autopano default settings. We are detecting control point on the same level / inside a stack, whereas ptgui cannot and just do hard linked stacks.
It often really changes the results.

no avatar
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Mon Aug 05, 2013 5:42 pm

AlexandreJ wrote:For bracketed series, it depends of autopano default settings. We are detecting control point on the same level / inside a stack, whereas ptgui cannot and just do hard linked stacks.
It often really changes the results.

Alexandre,

At least the right man for the right answer....(me likes speaking to the CEO)....if I would have used stacks you would (of course) be right. Then you see that Kolor tries to blend each layer based on controlpoints on each level (which is smart since moving targets can verry well be in different places in different levels). Now I shoot an empty lobby....no body moving....not even an insect. So in theory I could do without this functionality. What's even more I use Klaus's method....do Exposure blending first and merge later (so I used Photomatix first and created a set of exposure blended images). Then I took those and fed them to Kolor....no stacking there since well all it had were 7 gifs (6 60 degrees -5 + zenith)...and yet it ghosted, not as strong and convincingly as when using Kolor to do the whole of the stacking, but I had to do a lot of toughup work.....

But see for yourself....I could include the PTGui version but believe me that is spotless.....maybe some of the young bright things at your place should look into this, since this behaviour is strange as hell....no reason why it should do that.....if I take the middle set from the original stack no problem at all.....


Last edited by Artisan New on Mon Aug 05, 2013 5:45 pm, edited 1 time in total.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

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

by mediavets » Mon Aug 05, 2013 6:12 pm

AlexandreJ wrote:For bracketed series, it depends of autopano default settings. We are detecting control point on the same level / inside a stack, whereas ptgui cannot and just do hard linked stacks.
It often really changes the results.

Alex,

The image set I am using as an example here is a set of images already exposure-fused using Photomatix - so only one image per position.

So Detecting within stacks is not relevant in this case to any explanation of why PTGui Pro 9.17 does so much better a job than APG 3.x.
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
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Mon Aug 05, 2013 6:47 pm

You shoot a lobby using bracketing. Right?

Correct,

Then you used "my" method - it´s not "my" method but one method i use among others - of processing the bracketed images first and feed them to APG s "7 gifs" . . . GIFs? Are you sure?

Incorrect meant tif....16 bit and all....sorry, 34 degrees is not room temp to me.

What do you mean "yet it ghosted"?

See for yourself Klaus, on the picture ghosts of the vertical rod of the panohead are visible....I think....or I have to call Discovery channel...and some ghosthunters.

"if i take the middle set from the original stack" . . what does that mean? Didn´t you say you didn´t feed APG no stack at all?

(-2EV)-(-1,333EV)-(-0,666EV)-(0EV)-(+0.666EV)-(+1.333EV)-(+2EV)
7 pictures in a stack....take the middle one (0EV), do that for all 7 pictures that compose the sphere, put them (and only them as if they were shot not using HDR) into Autopano Pro 2.6.4. and let it compute.....NO PROBLEMS......then merge the stack in Photomatix to 7 TIF's, put them into Autopano Pro 2.6.4. and let it compute, GHOSTS.

Greets, Ed.
Last edited by Artisan New on Mon Aug 05, 2013 6:50 pm, edited 1 time in total.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

no avatar
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Mon Aug 05, 2013 6:48 pm

Indeed Andrew,

As I already explained in more words (to many for some I'm affraid, my fault entirely i guess :)),

greets, Ed.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

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

by mediavets » Mon Aug 05, 2013 7:10 pm

lumelix wrote:Hi
I think the colors are much better in the APG-example.
Stitching around the nadir with it's moving panohead-parts is something fatefully and differs from case to case.

Ignore the colours...

I decided to see what happened if I went right back to APG 2.0.9, the last version before the new blender and when Smartblend was still available, from 2010.

Observations:

1. I had forgotten how slow APG 2.0.9 was to render with Smartblend.

2. But.... the stitch was perfect with default settings, unlike APG 3.0.7

3. And it blended in a much more attractive way at the nadir than with APG 3.0.7 - see below.

Conclusion - in this instance other than render speed it would appear that APG 3.0.7 is a step backwards from APG 2.0.9.

Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.


Last edited by mediavets on Mon Aug 05, 2013 7:16 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.

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 Aug 05, 2013 7:56 pm

Artisan New wrote:(-2EV)-(-1,333EV)-(-0,666EV)-(0EV)-(+0.666EV)-(+1.333EV)-(+2EV)

Sorry - i don´t get this . . am confused by this funny EV numbers i guess . . :cool:

+/-2EV means: -2EV/0/+2EV. Which means 3 shots total. You show 5 shots total.

Again: +/- 2EV would be in speed-values: 1/8, 1/30, 1/125 sec. 3 steps. Not 5 steps.

The values between the steps must be equal. You need to tell Photomatix:
the number of steps (3 or 5) and the EV-value between the steps (2 or 1).

With 5 steps +/- 1 EV that would be: 1/8, 1/15, *1/30*, 1/60, 1/125 - with 1/30 as the center-value.

5 steps with a 1EV value most likely will be smoother than 3 steps with 2EV.

And so on.

Why you do have unequal EV values is a kid of mistery to me . .

+/- 2 EV does NOT mean the NUMBER of shots - it means the VALUE of the EV-DISTANCE.

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

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 Aug 05, 2013 8:06 pm

mediavets wrote:Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.

Here you ca see very nicely why i avoid shooting straight downward and seeing the heads base . . . . :D
Leave a hole instead and the ghosts are gone.

With a fisheye I use to shoot the center-row @-12° + Zenith. Works perfectly.
With a 35mm i shoot the lower row -12°. Leaves a very small hole and NO ghosts. Never.

This way i have only a very small fraction of the head in the image - which never confused CPs or blending so far.
This ghost-phenomenon i never have these days. I had it sometimes years ago - when i used my manual head the wrong way ;)

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

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

by mediavets » Mon Aug 05, 2013 8:09 pm

klausesser wrote:
mediavets wrote:Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.

Here you can see very nicely why i avoid shooting straight downward and seeing the heads base . . . . :D
Leave a hole instead and the ghosts are gone.

These are not stitched from images I shot. In fact the camera was not pointed straight down. But I know what you mean and would myself choose to avoid having the rotator in shot.

With a fisheye I use to shoot the center-row @-12° + Zenith. Works perfectly.
With a 35mm i shoot the lower row -12°. Leaves a very small hole and NO ghosts. Never.

Is that what the Panoneed automatically computes?
Last edited by mediavets on Mon Aug 05, 2013 8:15 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.

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 Aug 05, 2013 8:26 pm

mediavets wrote:
With a fisheye I use to shoot the center-row @-12° + Zenith. Works perfectly.
With a 35mm i shoot the lower row -12°. Leaves a very small hole and NO ghosts. Never.

Is that what the Panoneed automatically computes?

Yes - that´s how i had Josef program it for me. Independent from the focal-length: i always have the lower row set the way the lens sees only
a small fraction of the horizontal rail.

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

no avatar
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Tue Aug 06, 2013 6:47 am

I shot -2 till 2 EV using 2/3 EV steps.....now if you don't understand why....well my camera at the time could shoot either 1/3 or 1/2 EV step and either 3, 5 or 7 brackets....now if you think that is nonsence, I know but it means I had to use 7 steps to have 4 EV range, so if you want to you can take it up with Panasonic, my new Olympus can do 5 steps of 1 EV spanning 4 EV and the new Pen can span 6EV in 7 steps. But of course you're right I can also use only the -2EV, the 0EV and the 2EV position. I have no clue (eh, except maybe shutter wear) why camera companies don't integrate at least a 2 EV step and at least 5 steps in their camera's, but most of them simply don't. But all in all that is irrevant as Andrew's Kolor archeology shows....reason enough to postpone upgrade and to go for crossgrade and work with both, thanks Andrew.

Greets, Ed.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

no avatar
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Tue Aug 06, 2013 6:56 am

klausesser wrote:
mediavets wrote:
With a fisheye I use to shoot the center-row @-12° + Zenith. Works perfectly.
With a 35mm i shoot the lower row -12°. Leaves a very small hole and NO ghosts. Never.

Is that what the Panoneed automatically computes?

Yes - that´s how i had Josef program it for me. Independent from the focal-length: i always have the lower row set the way the lens sees only
a small fraction of the horizontal rail.

best, Klaus

Unfortunatly with my Panosaurus that was not possible.....so Kolor should put a warning on its site that Kolor Autopano only works with selected heads.....not with every head :) and they should appoint Klaus to be there head tester. And if you want to use a head not certified by Klaus you should use PTGui instead :). PTGui will have no objections to that.....:lol:

But seriously that could be the case Klaus......we are getting somewhere......and when you have enough contrast in the photo (like in a non exposure blended one) things can be corrected, but when the contrast drops, the new algorithm (though faster) hits it limits and won't be able to remove the ghost of the right arm of the head. And since PTGui is using a different algorithm it tolerates more (and is in this aspect) the better progam, clearly and deffinately.

Greets, Ed.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

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 Aug 06, 2013 9:19 am

mediavets wrote:1. I had forgotten how slow APG 2.0.9 was to render with Smartblend.
2. But.... the stitch was perfect with default settings, unlike APG 3.0.7
3. And it blended in a much more attractive way at the nadir than with APG 3.0.7 - see below.

Conclusion - in this instance other than render speed it would appear that APG 3.0.7 is a step backwards from APG 2.0.9.
Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.

It seems that you referring a lot to the way each anti-ghost algorithm handles all images. Approach are not the same at all:
1 -> Smartblend is doing two images comparison approach. It takes 2 images, solve the ghost issue, store the result. Take a new image, solve the ghost between the previous result and the new image, store the results. Do it until no more image. It is clearly not the right approach as you can come up with several 3 images panoramas that cannot be solved by smartblend.
3 -> PtGui, I don't know exactly the internal part, but it seems here that the approach is to keep at least something from each image. It gives you that strange tripod look where a small slice is coming from each image.
2 -> In autopano, we do antighost globally. We try to find the zones that will give you the less noticeable seam. On the nadir, this algorithm can decide to keep one image fully because the others don't really fit in this "less noticeable energy". So the result is the one you have on this case. It doesn't seem for me contradictory to the other results, worth or better.

Now what can be done ?

With our global approach, we can easily add some new energies that says : keep me each images, try to keep the center, try to make a symmetric nadir, etc.
We did it before with the focal length energy. If you blend together several focal length images, you can decide if the ghost have priority or the focal length ( preserving seams invisibility or preserving the pixel with highest density ).

Should we do something ?

I don't know. For me this tripod should not be here and everything that helps to remove it quickly is for me highest priority.

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 » Tue Aug 06, 2013 10:54 am

AlexandreJ wrote:I don't know. For me this tripod should not be here and everything that helps to remove it quickly is for me highest priority.

Using the masking tool can be a solution in this case ? Or Klaus suggestion, not to shoot the rotator/panohead that moves around.
Regards
Martin

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

by mediavets » Tue Aug 06, 2013 12:27 pm

AlexandreJ wrote:
mediavets wrote:1. I had forgotten how slow APG 2.0.9 was to render with Smartblend.
2. But.... the stitch was perfect with default settings, unlike APG 3.0.7
3. And it blended in a much more attractive way at the nadir than with APG 3.0.7 - see below.

Conclusion - in this instance other than render speed it would appear that APG 3.0.7 is a step backwards from APG 2.0.9.
Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.

It seems that you referring a lot to the way each anti-ghost algorithm handles all images.

Alex,

The nadir area was one (obvious) example of the differences resulting from using APG 2.0.9, APG 3.0.7 and PTGui Pro 9.17.

Another area which showed a big difference was the image appearing on the LCD monitor on the desk.

PTGui Pro and APG 2.0.9 showed a similar image; APG 3.0.7 showed a completeley different image.

And most importantly both APG 2.0.9 and PTGui Pro 0.17 produced an apparently perfect stitch with default settings.

On the other hand APG 3.0.7 produced a far from perfect stitch and I was unable fully to resolve the stitch errors even with extensive control point editing. The stitched pano looks fine in the Editor preview but when rendered there are significant errors. This subject has been raised elsewhere and to date there doesn't seem to be an entirely satisfactory 'answer'/explanation.

I noticed that PTGUi Pro seemed to auto detect CPs over a wider 'spread'/area in the overlaps, whereas APG 3.0.7 seemed to cluster CPs around the 'equator' of linked pairs of images even though there were excellent CP targets elsewhere along the overlap.

Some users have the impression that APG 3.x doesn't handle fused exposure images produced by Photomatix as well as it handles single exposure images, nor as well as APG 2.0.9 and PTGui Pro handle fused exposure images from Photomatix. Might this be the case?

The quality of the rendered image seemed much better with APG 2.0.9 and PTGui Pro 9.17 too, compared to the output from APG 3.0.7.

Is this all the result of the new blender and rendering engine in APG 3.x?
Last edited by mediavets on Tue Aug 06, 2013 12:50 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
Artisan New
Member
 
Posts: 174
Likes: 0 post
Liked in: 0 post
Joined: Thu Jul 25, 2013 8:23 am
Info

by Artisan New » Tue Aug 06, 2013 3:13 pm

I don't know. For me this tripod should not be here and everything that helps to remove it quickly is for me highest priority.

Do so Alexandre.....APG/APP should benefit from it. [DELETED CONTENT]

Greets, Ed.
Olympus OM-D with HLD-6, Fuji GX680, Samyang 7.5, Olympus 9-18, Sigma 19, Panasonic 14-45, Nikon 50 1.4 on Novoflex with tripod mount, Nikon 80-200, Panasonic 45-200, Fujinon 135, Fujinon 80, Fujinon 65, Fujinon 50, Gitzo Gilux Reporter 2, Sirui Ballhead and Panosaurus 2.0 NNP adapter, Motorized Panohead of Canadian (ironware)/Dutch (electronics and software) construction.

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 Aug 06, 2013 4:31 pm

AlexandreJ wrote:
mediavets wrote:1. I had forgotten how slow APG 2.0.9 was to render with Smartblend.
2. But.... the stitch was perfect with default settings, unlike APG 3.0.7
3. And it blended in a much more attractive way at the nadir than with APG 3.0.7 - see below.

Conclusion - in this instance other than render speed it would appear that APG 3.0.7 is a step backwards from APG 2.0.9.
Screenshots of the nadir area from the left - APG 2.0.9 - APG 3.0.7- PTGui Pro 9.17.

It seems that you referring a lot to the way each anti-ghost algorithm handles all images. Approach are not the same at all:
1 -> Smartblend is doing two images comparison approach. It takes 2 images, solve the ghost issue, store the result. Take a new image, solve the ghost between the previous result and the new image, store the results. Do it until no more image. It is clearly not the right approach as you can come up with several 3 images panoramas that cannot be solved by smartblend.
3 -> PtGui, I don't know exactly the internal part, but it seems here that the approach is to keep at least something from each image. It gives you that strange tripod look where a small slice is coming from each image.
2 -> In autopano, we do antighost globally. We try to find the zones that will give you the less noticeable seam. On the nadir, this algorithm can decide to keep one image fully because the others don't really fit in this "less noticeable energy". So the result is the one you have on this case. It doesn't seem for me contradictory to the other results, worth or better.

Now what can be done ?

With our global approach, we can easily add some new energies that says : keep me each images, try to keep the center, try to make a symmetric nadir, etc.
We did it before with the focal length energy. If you blend together several focal length images, you can decide if the ghost have priority or the focal length ( preserving seams invisibility or preserving the pixel with highest density ).

Should we do something ?

I don't know. For me this tripod should not be here and everything that helps to remove it quickly is for me highest priority.

Hi Alexandre!

I never succeded to use stacks for a good stitch in the past. That was before 3.0.7.

Some days ago i gave it another try and was surprised how good it worked. Sometimes it simply is preferable being able to do a bracketed stitch very fast . . . on-scene for getting a reliable feel what it´ll be about in the end.

Some words to smartblend or not: i´m very glad you abandoned smartblend. It was very slow and even more hardware-dependent than the actual version of multiblend. But most important: smartblend tended to "smear" areas of overlap. I didn´t encouter it since then.

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

Next

Who is online

Users browsing this forum: No registered users and 3 guests

cron