Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 6 Jul 2008 at 21:16 GMT
|
reply
|
Hello everyone 
I know this is my first post in this forum but I have been reading it for weeks now trying to grasp as much information as possible. When I am on a quest to do something that needs learning, I try as much as possible to search and read everything I come across anywhere.
I have been taking cylindrical panos for a while now but mostly of object far away where parallax errors have no visible effects on the end results. But I tried to take a couple of panos of close objects and parallax errors were major. To tell you the truth before I came across these errors I didnt know about them or anything in regards to proper camera/lens rotation and NP points & anything related to that! So I started digging the net for information about this and why I had the errors (I learned photography pretty much from the net). I realized I needed a good pano head and I settled on Nodal Ninja for its value for the money and its great reviews. I found my camera/lens combination numbers listed under the compatibility list for NN3 MKII. I now realize I could have done better by getting the NN5, but what is done is done! the smaller the equipment the better since I mostly photograph during travel.
I got the NN3 a couple of weeks ago and started playing with it and searching the net for the best set up for my camera/lens to be able to create 360x180 panos. That was a hard thing to do since not many VR photographers use this comb. for panos. But I came across a couple of recommendations which I logged and tried each one of them and finally settled on the Josh Sommers setup, but with some issues that I will share with you later. Every setup could be correct since someone mentioned it online (which most likely means they've tried it!) but not every setup is necessarily optimum for everyone!
apologies for the long introduction but I had to start this thread properly
------------------
My equipment:
Canon 40 D
EFS 10-22 mm
Manfrotto 190XPRO tripod
Manfrotto 2way bubble level
NN3 MKII
------------------
Trials & results:
1- 6 @ -30, 6 @ +30, N, Z (merge errors)
2- 8 @ -45, 8 @ +45, N (merge errors)
3- 4 @ -45, 8 @ 0, 4 @ +45 (minor merge errors)
by now I got frustrated  I decided to remove the NN camera plate and purchased a 2 way leveler to make sure i was rotating the camera correctly
4- 4 @ -45, 8 @ 0, 4 @ +45, N (second test: blending failed!! messed up result)
5- 8 @ -30, 8 @ +30, N, Z ( good stitch results but definitely need Z & N shots)
6- 8 @-30, 8 @ +45, N (J.S. setup: worked like a charm  )
-----
I did a couple of tests with setup 6, & found the following:
- you dont need N shot if you are going to use a mirror.
- N zone has small footprint and can easily be covered us PS 
- you will need to take N shot if you want high quality end result
- no need for Z shot unless ...
- if you are shooting HDR and the Z zone have elements of varying exposures you will need to take Z shots to ensure high quality area or you can settle for PS editing! 
----------------
Now for the result!
this is the latest test I did, I am saying test because I know this pano is not of great photographic quality and has some errors. In addition, I took it in my back yard without any photographic subject around. It is hard to take outdoor photographs in Saudi Arabia these days for multiple reasons ( extreme hot weather 45-50 C during day, sky almost grey or dull blue without any cloud in sight, that is if there is no sand storm or dusty sky!!!) lol 
HDR, 16 shots x 3 exposures with PS edits to cover N and correct the Z blending.
I got minor blending errors in 2 or 3 spots (look in the movie) I dont know why I have these errors. Is this normal? Do I have to cope with such errors in every VR i will do and correct it in post processing? any ideas?
the movie around 470 k:
www.al-riyadh.org/360/BackYard/360x180-hrd.html
the planet:
I hope my experience will benefit someone out there who has the same equipment as I. I also hope to find an answer/possible reasons to my blending errors here.
regards,
Ayman
www.al-riyadh.org/English/
|
|
alert moderator
|
|
Steven Scott
Posts: 38
Location: Lahaina, Maui, Hawaii, United States
Registered: 13 Mar 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 1:16 GMT
|
reply
|
Ayman,
I have no answers, but will watch comments/suggestions very carefully. I plan to purchase very soon, the Canon 10-22mm lens and either the Canon 40D or the Canon XSi (450). I have access to both -- have done several tests -- and am having problems, too.
Good luck --- and I'll await some suggestions, too!
mahalo, steve
|
|
alert moderator
|
|
John Houghton
Posts: 2265
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 6:51 GMT
|
reply
|
Ayman, You have given lots of detail about your equipment and shooting tests but no information about the stitching, blending and HDR software you are using and how it is used. If your pano head setup is correct, then I would not expect stitching problems with any of your shooting configurations. But not all stitching programs are equally accurate or equally easy to use. You mention a 2-way leveller, but that is completely unnecessary and should not have any effect on the stitching quality.
All the blending programs (such as Enblend and Smartblend) have problems blending at the nadir and zenith, which correspond to the areas at the top and bottom of the 360x180 equirectangular panorama image. Zenith and nadir shots will generally avoid such problems, though you should ensure there are holes to be filled by the Z and N shots, otherwise the blender might discard them as being surplus to requirements.
HDR blending is more problematical. It can easily produce obvious blending artifacts, particularly in the neighborhood of sharp transitions of brightness. The blending needs to be carefully controlled to avoid unnatural overall colour and contrast effects.
John
|
|
alert moderator
|
|
mediavets
Posts: 419
Location: Isleham, Cambs., United Kingdom
Registered: 8 Feb 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 8:03 GMT
updated: 7 Jul 2008 at 8:04 GMT
|
reply
|
John Houghton said: HDR blending is more problematical. It can easily produce obvious blending artifacts, particularly in the neighborhood of sharp transitions of brightness. The blending needs to be carefully controlled to avoid unnatural overall colour and contrast effects.
John
Yes; I agree, perhaps you should forget about HDR stuff until you have everything else sorted to your satisfaction.
Andrew
|
|
alert moderator
|
|
Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 12:12 GMT
|
reply
|
Steven Scott said: and I'll await some suggestions, too!
Thanks steve, we are here to learn and share knowledge so I hope this thread would be of benefit.
John Houghton said: Ayman, You have given lots of detail about your equipment and shooting tests but no information about the stitching, blending and HDR software you are using and how it is used.
Thanks John for your reply. I didnt think it would make a big difference to mention the software... but now that you asked, I use Autopano for stitching and photomatix for blending. My stitching errors not really related to HDR I only tried it for the sake of trial and to see how it works, the same errors in blending existed in the none HDR and the HDR versions. I mentioned HDR here to share my findings related to the final result and mainly the Z shot. I agree with you and Andrew of the HDR being more problematic to begin with, but this is not the fist time I do HDR pano blending, but it is the first time i do it for a full 360x180.
I took 3 different exposures per shot, batch blended them in photomatix then merged the finals in autopano.
I tried as much as possible to set up the head correctly as per the numbers listed in the site for the NP (EP) with the correct plates and all, I used an online calculator to measure the coverage of the FOV between shots and it gave me a 39% overlap for 8 shots which I think is more than enough. I also took all shots using remote shutter to minimize any shake.
The reason I got the leveler is just to make sure that I was leveled through out the rotation even though it wouldnt make a big difference, but just to make sure, and besides it is a useful piece of hardware to have
I realize now that I should be more specific to the errors I faced, unfortunately I am not using my personal machine right now, but I will post another reply later on tonight when I get home with specific shot pieces I faced problems with which would give you and others a much better understanding of what I am facing.
regards,
Ayman
|
|
alert moderator
|
|
mediavets
Posts: 419
Location: Isleham, Cambs., United Kingdom
Registered: 8 Feb 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 12:31 GMT
updated: 7 Jul 2008 at 12:32 GMT
|
reply
|
Ayman said:
The reason I got the leveler is just to make sure that I was leveled through out the rotation even though it wouldnt make a big difference, but just to make sure, and besides it is a useful piece of hardware to have
Ayman
You should set the level for first shot and NOT change it during rotation (regardless of what you see happening with your bubble level).
Levelling really doesn't matter if you are shooting 360x180 as you can always level the stitched pano in your stitching software. It's particularly easy with Autopano Pro, there's video tutorial if you are not sure how to do it:
www.autopano.net/wiki/action/view/How_to_straight...
Andrew
|
|
alert moderator
|
|
Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 12:49 GMT
|
reply
|
mediavets said:
You should set the level for first shot and NOT change it during rotation (regardless of what you see happening with your bubble level).
Levelling really doesn't matter if you are shooting 360x180 as you can always level the stitched pano in your stitching software. It's particularly easy with Autopano Pro, there's video tutorial if you are not sure how to do it:
www.autopano.net/wiki/action/view/How_to_straight...
Andrew
you are absolutely right. After I set my gear up and point my camera towards the first shot, I start by rotating the camera around the view for all 8 stops before i take any shots to make sure my level doesn't change during rotation. I know this could be an exaggeration and I know I could always set the level in APP but i wanted to minimize any error factors at least in the beginning until I gain enough experience and done enough VRs to know what may and may not affect my final result.
Ayman
|
|
alert moderator
|
|
Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 18:29 GMT
|
reply
|
Hi again,
here is a compilation of the errors I faced in the pano merging. I took them from the non-hdr version since it is clearer there with no artifacts (the other one had many compression artifacts). Not exactly the same errors, but they are around the same areas and with almost the same small amount.
|
|
alert moderator
|
|
mediavets
Posts: 419
Location: Isleham, Cambs., United Kingdom
Registered: 8 Feb 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 19:02 GMT
updated: 7 Jul 2008 at 19:03 GMT
|
reply
|
Ayman said: mediavets said:
You should set the level for first shot and NOT change it during rotation (regardless of what you see happening with your bubble level).
Levelling really doesn't matter if you are shooting 360x180 as you can always level the stitched pano in your stitching software. It's particularly easy with Autopano Pro, there's video tutorial if you are not sure how to do it:
www.autopano.net/wiki/action/view/How_to_straight...
Andrew
you are absolutely right. After I set my gear up and point my camera towards the first shot, I start by rotating the camera around the view for all 8 stops before i take any shots to make sure my level doesn't change during rotation. I know this could be an exaggeration and I know I could always set the level in APP but i wanted to minimize any error factors at least in the beginning until I gain enough experience and done enough VRs to know what may and may not affect my final result.
Ayman
With an NN3 - in fact on most if not all pano heads - the bubble in the level will move as you rotate the head; it doesn't matter and you should not seek to adjust the level between shots.
Andrew
|
|
alert moderator
|
|
Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 19:18 GMT
|
reply
|
mediavets said: With an NN3 - in fact on most if not all pano heads - the bubble in the level will move as you rotate the head; it doesn't matter and you should not seek to adjust the level between shots.
Andrew
true, I read the same comment before. I dont try to adjust based on the bubble level in the head itself. I only adjust once based on the 2way bubble level attached to the camera. I just rotate around to make sure I got the right adjustments without readjustments between shots.
Ayman
|
|
alert moderator
|
|
mediavets
Posts: 419
Location: Isleham, Cambs., United Kingdom
Registered: 8 Feb 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 20:05 GMT
updated: 7 Jul 2008 at 20:06 GMT
|
reply
|
Ayman said: Hi again,
here is a compilation of the errors I faced in the pano merging. I took them from the non-hdr version since it is clearer there with no artifacts (the other one had many compression artifacts). Not exactly the same errors, but they are around the same areas and with almost the same small amount.
If you wish to put the image set on-line somewhere where I can download it I'd be happy yo have a go at stitching it with APP to see if I can help figure out your problem.
Hopefully you will not have pre-processed the images and stripped out the EXIF data. I'd prefer JPEGs as I have no tools to convert Canon RAWs being a Nikon shooter.
Andrew
|
|
alert moderator
|
|
Ayman
Posts: 23
Location: Dhahran, Saudi Arabia
Registered: 5 Jul 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 7 Jul 2008 at 20:27 GMT
|
reply
|
great. I would be happy to upload it, just give me some time to get hold of a connection with fast upload speeds. I am taking shots with med quality JPEGs and no pre processing since this is only intended for testing, so no problems there.
regards,
Ayman
|
|
alert moderator
|
|
mediavets
Posts: 419
Location: Isleham, Cambs., United Kingdom
Registered: 8 Feb 2008
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 8 Jul 2008 at 8:10 GMT
updated: 8 Jul 2008 at 8:20 GMT
|
reply
|
OK - in the meantime it would be good to see a screenshot of your pano in the Control Points Editor.
www.autopano.net/wiki/action/view/Control_Points_...
Your pano scene has many relatively plain areas with repeating patterns - the blockwork walls of the compound, the paved ground surface and the slatted overhead sunscreens - and the automatic control points detection process of APP can have difficulties with this sort of thing. The branches of the tree moving in the wind could also cause problems for automatic control point detection.
Use the Control Points Editor view to check whether APP automatically created redundant, unsuitable or incorrect links between images and correct/delete as appropriate. Then optimise and try rendering again.
Which interpolator and blender are you using when rendering the final pano image?
www.autopano.net/wiki/action/view/Interpolation_a...
www.autopano.net/wiki/action/view/Render_Settings
Since this thread seems to be becoming a technical discourse about using Autopano Pro perhaps it would be better to continue the discussion on the APP forum?
www.autopano.net/forum/
For one thing it is much easier to post images, screenshots, etc on the APP forum. And you will find more APP expertise there than on the Panoguide forum where most members are PTGui users.
Andrew
|
|
alert moderator
|
|
Matt Rogers
[360 Precision]
Posts: 209
Location: Oxford UK, United Kingdom
Registered: 16 Jun 2005
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 8 Jul 2008 at 8:49 GMT
|
reply
|
Quite simply your control points are not good enough or are not wide spread enough across the images or both. A part of the problem is PTGui will report good controls when in-fact they're not really good at all. Unless PTGui reports "too good to be true" you need to go back and either add more control points or delete the worse control points and re-optimise.
You also have to remember that the control point distance is directly related to your final output size. I often see people optimise the control points whilst the output size is still relatively small, 2800 x 1400. Doing this will only give you false positives unless you either output at this size OR change the output size and THEN re-optimise.
The errors above seem to be in the region of 2-3 pixels so I'm guessing your average CP distance is in the region. To remove these errors you really want your control points all be below 1px. If this isn't possible then you might not have the NN3 set-up correctly for your camera and lens.
Matt
|
|
alert moderator
|
|
John Houghton
Posts: 2265
Location: Hitchin, United Kingdom
Registered: 17 Jan 2005
|
Re: Canon 40D/ EFS 10-22/ NN3 MKII - trials & results - questions too!
Posted: 8 Jul 2008 at 9:17 GMT
|
reply
|
Matt Rogers said: You also have to remember that the control point distance is directly related to your final output size. I often see people optimise the control points whilst the output size is still relatively small, 2800 x 1400. Doing this will only give you false positives unless you either output at this size OR change the output size and THEN re-optimise.
Note that this only applies when using the Panorama Tools optimizer. Most people, particularly novices, will be using the PTGui optimizer, which reports cp distances relative to the resolution of the input images.
John
|
|
alert moderator
|
|
Prev
